The Smart Cyber Ecosystem for Sustainable Development: Principles, Building Blocks, and Paradigms [1 ed.] 1119761646, 9781119761648

Table of contents :
Cover
Half-Title Page
Series Page
Title Page
Copyright Page
Contents
Preface
Part 1: INTERNET OF THINGS
1 Voyage of Internet of Things in the Ocean of Technology
1.1 Introduction
1.1.1 Characteristics of IoT
1.1.2 IoT Architecture
1.1.3 Merits and Demerits of IoT
1.2 Technological Evolution Toward IoT
1.3 IoT-Associated Technology
1.4 Interoperability in IoT
1.5 Programming Technologies in IoT
1.5.1 Arduino
1.5.2 Raspberry Pi
1.5.3 Python
1.6 IoT Applications
Conclusion
References
2 AI for Wireless Network Optimization: Challenges and Opportunities
2.1 Introduction to AI
2.2 Self-Organizing Networks
2.2.1 Operation Principle of Self-Organizing Networks
2.2.2 Self-Configuration
2.2.3 Self-Optimization
2.2.4 Self-Healing
2.2.5 Key Performance Indicators
2.2.6 SON Functions
2.3 Cognitive Networks
2.4 Introduction to Machine Learning
2.4.1 ML Types
2.4.2 Components of ML Algorithms
2.4.3 How do Machines Learn?
2.4.4 ML and Wireless Networks
2.5 Software-Defined Networks
2.5.1 SDN Architecture
2.5.2 The OpenFlow Protocol
2.5.3 SDN and ML
2.6 Cognitive Radio Networks
2.6.1 Sensing Methods
2.7 ML for Wireless Networks: Challenges and Solution Approaches
2.7.1 Cellular Networks
2.7.2 Wireless Local Area Networks
2.7.3 Cognitive Radio Networks
References
3 An Overview on Internet of Things (IoT) Segments and Technologies
3.1 Introduction
3.2 Features of IoT
3.3 IoT Sensor Devices
3.4 IoT Architecture
3.5 Challenges and Issues in IoT
3.6 Future Opportunities in IoT
3.7 Discussion
3.8 Conclusion
References
4 The Technological Shift: AI in Big Data and IoT
4.1 Introduction
4.2 Artificial Intelligence
4.2.1 Machine Learning
4.2.2 Further Development in the Domain of Artificial Intelligence
4.2.3 Programming Languages for Artificial Intelligence
4.2.4 Outcomes of Artificial Intelligence
4.3 Big Data
4.3.1 Artificial Intelligence Methods for Big Data
4.3.2 Industry Perspective of Big Data
4.4 Internet of Things
4.4.1 Interconnection of IoT With AoT
4.4.2 Difference Between IIoT and IoT
4.4.3 Industrial Approach for IoT
4.5 Technical Shift in AI, Big Data, and IoT
4.5.1 Industries Shifting to AI-Enabled Big Data Analytics
4.5.2 Industries Shifting to AI-Powered IoT Devices
4.5.3 Statistical Data of These Shifts
4.6 Conclusion
References
5 IoT’s Data Processing Using Spark
5.1 Introduction
5.2 Introduction to Apache Spark
5.2.1 Advantages of Apache Spark
5.2.2 Apache Spark’s Components
5.3 Apache Hadoop MapReduce
5.3.1 Limitations of MapReduce
5.4 Resilient Distributed Dataset (RDD)
5.4.1 Features and Limitations of RDDs
5.5 DataFrames
5.6 Datasets
5.7 Introduction to Spark SQL
5.7.1 Spark SQL Architecture
5.7.2 Spark SQL Libraries
5.8 SQL Context Class in Spark
5.9 Creating DataFrames
5.9.1 Operations on DataFrames
5.10 Aggregations
5.11 Running SQL Queries on DataFrames
5.12 Integration With RDDs
5.12.1 Inferring the Schema Using Reflection
5.12.2 Specifying the Schema Programmatically
5.13 Data Sources
5.13.1 JSON Datasets
5.13.2 Hive Tables
5.13.3 Parquet Files
5.14 Operations on Data Sources
5.15 Industrial Applications
5.16 Conclusion
References
6 SE-TEM: Simple and Efficient Trust Evaluation Model for WSNs
6.1 Introduction
6.1.1 Components of WSNs
6.1.2 Trust
6.1.3 Major Contribution
6.2 Related Work
6.3 Network Topology and Assumptions
6.4 Proposed Trust Model
6.4.1 CM to CM (Direct) Trust Evaluation Scheme
6.4.2 CM to CM Peer Recommendation (Indirect) Trust Estimation (PR
(.t))
6.4.3 CH-to-CH Direct Trust Estimation
6.4.4 BS-to-CH Feedback Trust Calculation
6.5 Result and Analysis
6.5.1 Severity Analysis
6.5.2 Malicious Node Detection
6.6 Conclusion and Future Work
References
7 Smart Applications of IoT
7.1 Introduction
7.2 Background
7.2.1 Enabling Technologies for Building Intelligent Infrastructure
7.3 Smart City
7.3.1 Benefits of a Smart City
7.3.2 Smart City Ecosystem
7.3.3 Challenges in Smart Cities
7.4 Smart Healthcare
7.4.1 Smart Healthcare Applications
7.4.2 Challenges in Healthcare
7.5 Smart Agriculture
7.5.1 Environment Agriculture Controlling
7.5.2 Advantages
7.5.3 Challenges
7.6 Smart Industries
7.6.1 Advantages
7.6.2 Challenges
7.7 Future Research Directions
7.8 Conclusions
References
8 Sensor-Based Irrigation System: Introducing Technology in Agriculture
8.1 Introduction
8.1.1 Technology in Agriculture
8.1.2 Use and Need for Low-Cost Technology in Agriculture
8.2 Proposed System
8.3 Flow Chart
8.4 Use Case
8.5 System Modules
8.5.1 Raspberry Pi
8.5.2 Arduino Uno
8.5.3 DHT 11 Humidity and Temperature Sensor
8.5.4 Soil Moisture Sensor
8.5.5 Solenoid Valve
8.5.6 Drip Irrigation Kit
8.5.7 433 MHz RF Module
8.5.8 Mobile Application
8.5.9 Testing Phase
8.6 Limitations
8.7 Suggestions
8.8 Future Scope
8.9 Conclusion
Acknowledgement
References
Suggested Additional Readings
Key Terms and Definitions
Appendix
Example Code
9 Artificial Intelligence: An Imaginary World of Machine
9.1 The Dawn of Artificial Intelligence
9.2 Introduction
9.3 Components of AI
9.3.1 Machine Reasoning
9.3.2 Natural Language Processing
9.3.3 Automated Planning
9.3.4 Machine Learning
9.4 Types of Artificial Intelligence
9.4.1 Artificial Narrow Intelligence
9.4.2 Artificial General Intelligence
9.4.3 Artificial Super Intelligence
9.5 Application Area of AI
9.6 Challenges in Artificial Intelligence
9.7 Future Trends in Artificial Intelligence
9.8 Practical Implementation of AI Application
References
10 Impact of Deep Learning Techniques in IoT
10.1 Introduction
10.2 Internet of Things
10.2.1 Characteristics of IoT
10.2.2 Architecture of IoT
10.3 Deep Learning
10.3.1 Models of Deep Learning
10.3.2 Applications of Deep Learning
10.3.3 Advantages of Deep Learning
10.3.4 Disadvantages of Deep Learning
10.3.5 Deployment of Deep Learning in IoT
10.3.6 Deep Learning Applications in IoT
10.3.7 Deep Learning Techniques on IoT Devices
10.4 IoT Challenges on Deep Learning and Future Directions
10.4.1 Lack of IoT Dataset
10.4.2 Pre-Processing
10.4.3 Challenges of 6V’s
10.4.4 Deep Learning Limitations
10.5 Future Directions of Deep Learning
10.5.1 IoT Mobile Data
10.5.2 Integrating Contextual Information
10.5.3 Online Resource Provisioning for IoT Analytics
10.5.4 Semi-Supervised Analytic Framework
10.5.5 Dependable and Reliable IoT Analytics
10.5.6 Self-Organizing Communication Networks
10.5.7 Emerging IoT Applications
10.6 Common Datasets for Deep Learning in IoT
10.7 Discussion
10.8 Conclusion
References
Part 2: ARTIFICIAL INTELLIGENCE IN HEALTHCARE
11 Non-Invasive Process for Analyzing Retinal Blood Vessels Using Deep Learning Techniques
11.1 Introduction
11.2 Existing Methods Review
11.3 Methodology
11.3.1 Architecture of Stride U-Net
11.3.2 Loss Function
11.4 Databases and Evaluation Metrics
11.4.1 CNN Implementation Details
11.5 Results and Analysis
11.5.1 Evaluation on DRIVE and STARE Databases
11.5.2 Comparative Analysis
11.6 Concluding Remarks
References
12 Existing Trends in Mental Health Based on IoT Applications: A Systematic Review
12.1 Introduction
12.2 Methodology
12.3 IoT in Mental Health
12.4 Mental Healthcare Applications and Services Based on IoT
12.5 Benefits of IoT in Mental Health
12.5.1 Reduction in Treatment Cost
12.5.2 Reduce Human Error
12.5.3 Remove Geographical Barriers
12.5.4 Less Paperwork and Documentation
12.5.5 Early Stage Detection of Chronic Disorders
12.5.6 Improved Drug Management
12.5.7 Speedy Medical Attention
12.5.8 Reliable Results of Treatment
12.6 Challenges in IoT-Based Mental Healthcare Applications
12.6.1 Scalability
12.6.2 Trust
12.6.3 Security and Privacy Issues
12.6.4 Interoperability Issues
12.6.5 Computational Limits
12.6.6 Memory Limitations
12.6.7 Communications Media
12.6.8 Devices Multiplicity
12.6.9 Standardization
12.6.10 IoT-Based Healthcare Platforms
12.6.11 Network Type
12.6.12 Quality of Service
12.7 Blockchain in IoT for Healthcare
12.8 Results and Discussion
12.9 Limitations of the Survey
12.10 Conclusion
References
13 Monitoring Technologies for Precision Health
13.1 Introduction
13.2 Applications of Monitoring Technologies
13.2.1 Everyday Life Activities
13.2.2 Sleeping and Stress
13.2.3 Breathing Patterns and Respiration
13.2.4 Energy and Caloric Consumption
13.2.5 Diabetes, Cardiac, and Cognitive Care
13.2.6 Disability and Rehabilitation
13.2.7 Pregnancy and Post-Procedural Care
13.3 Limitations
13.3.1 Quality of Data and Reliability
13.3.2 Safety, Privacy, and Legal Concerns
13.4 Future Insights
13.4.1 Consolidating Frameworks
13.4.2 Monitoring and Intervention
13.4.3 Research and Development
13.5 Conclusions
References
14 Impact of Artificial Intelligence in Cardiovascular Disease
14.1 Artificial Intelligence
14.2 Machine Learning
14.3 The Application of AI in CVD
14.3.1 Precision Medicine
14.3.2 Clinical Prediction
14.3.3 Cardiac Imaging Analysis
14.4 Future Prospect
14.5 PUAI and Novel Medical Mode
14.5.1 Phenomenon of PUAI
14.5.2 Novel Medical Model
14.6 Traditional Mode
14.6.1 Novel Medical Mode Plus PUAI
14.7 Representative Calculations of AI
14.8 Overview of Pipeline for Image-Based Machine Learning Diagnosis
References
15 Healthcare Transformation With Clinical Big Data Predictive Analytics
15.1 Introduction
15.1.1 Big Data in Health Sector
15.1.2 Data Structure Produced in Health Sectors
15.2 Big Data Challenges in Healthcare
15.2.1 Big Data in Computational Healthcare
15.2.2 Big Data Predictive Analytics in Healthcare
15.2.3 Big Data for Adapted Healthcare
15.3 Cloud Computing and Big Data in Healthcare
15.4 Big Data Healthcare and IoT
15.5 Wearable Devices for Patient Health Monitoring
15.6 Big Data and Industry 4.0
15.7 Conclusion
References
16 Computing Analysis of Yajna and Mantra Chanting as a Therapy: A Holistic Approach for All by Indian Continent Amidst Pandemic
16.1 Introduction
16.1.1 The Stats of Different Diseases, Comparative Observation on Symptoms, and Mortality Rate
16.1.2 Precautionary Guidelines Followed in Indian Continent
16.1.3 Spiritual Guidelines in Indian Society
16.1.4 Veda Vigyaan: Ancient Vedic Knowledge
16.1.5 Yagyopathy Researches, Say, Smoke of Yagya is Boon
16.1.6 The Yagya Samagri
16.2 Literature Survey
16.2.1 Technical Aspects of Yajna and Mantra Therapy
16.2.2 Mantra Chanting and Its Science
16.2.3 Yagya Medicine (Yagyopathy)
16.2.4 The Medicinal HavanSamagri Components
16.2.5 Scientific Benefits of Havan
16.3 Experimental Setup Protocols With Results
16.3.1 Subject Sample Distribution
16.3.2 Conclusion and Discussion Through Experimental Work
16.4 Future Scope and Limitations
16.5 Novelty
16.6 Recommendations
16.7 Applications of Yajna Therapy
16.8 Conclusions
Acknowledgement
References
Key Terms and Definitions
17 Extraction of Depression Symptoms From Social Networks
17.1 Introduction
17.1.1 Diagnosis and Treatments
17.2 Data Mining in Healthcare
17.2.1 Text Mining
17.3 Social Network Sites
17.4 Symptom Extraction Tool
17.4.1 Data Collection
17.4.2 Data Processing
17.4.3 Data Analysis
17.5 Sentiment Analysis
17.5.1 Emotion Analysis
17.5.2 Behavioral Analysis
17.6 Conclusion
References
Part 3: CYBERSECURITY
18 Fog Computing Perspective: Technical Trends, Security Practices, and Recommendations
18.1 Introduction
18.2 Characteristics of Fog Computing
18.3 Reference Architecture of Fog Computing
18.4 CISCO IOx Framework
18.5 Security Practices in CISCO IOx
18.5.1 Potential Attacks on IoT Architecture
18.5.2 Perception Layer (Sensing)
18.5.3 Network Layer
18.5.4 Service Layer (Support)
18.5.5 Application Layer (Interface)
18.6 Security Issues in Fog Computing
18.6.1 Virtualization Issues
18.6.2 Web Security Issues
18.6.3 Internal/External Communication Issues
18.6.4 Data Security Related Issues
18.6.5 Wireless Security Issues
18.6.6 Malware Protection
18.7 Machine Learning for Secure Fog Computing
18.7.1 Layer 1 Cloud
18.7.2 Layer 2 Fog Nodes For The Community
18.7.3 Layer 3 Fog Node for Their Neighborhood
18.7.4 Layer 4 Sensors
18.8 Existing Security Solution in Fog Computing
18.8.1 Privacy-Preserving in Fog Computing
18.8.2 Pseudocode for Privacy Preserving in Fog Computing
18.8.3 Pseudocode for Feature Extraction
18.8.4 Pseudocode for Adding Gaussian Noise to the Extracted Feature
18.8.5 Pseudocode for Encrypting Data
18.8.6 Pseudocode for Data Partitioning
18.8.7 Encryption Algorithms in Fog Computing
18.9 Recommendation and Future Enhancement
18.9.1 Data Encryption
18.9.2 Preventing from Cache Attacks
18.9.3 Network Monitoring
18.9.4 Malware Protection
18.9.5 Wireless Security
18.9.6 Secured Vehicular Network
18.9.7 Secure Multi-Tenancy
18.9.8 Backup and Recovery
18.9.9 Security with Performance
18.10 Conclusion
References
19 Cybersecurity and Privacy Fundamentals
19.1 Introduction
19.2 Historical Background and Evolution of Cyber Crime
19.3 Introduction to Cybersecurity
19.3.1 Application Security
19.3.2 Information Security
19.3.3 Recovery From Failure or Disaster
19.3.4 Network Security
19.4 Classification of Cyber Crimes
19.4.1 Internal Attacks
19.4.2 External Attacks
19.4.3 Unstructured Attack
19.4.4 Structured Attack
19.5 Reasons Behind Cyber Crime
19.5.1 Making Money
19.5.2 Gaining Financial Growth and Reputation
19.5.3 Revenge
19.5.4 For Making Fun
19.5.5 To Recognize
19.5.6 Business Analysis and Decision Making
19.6 Various Types of Cyber Crime
19.6.1 Cyber Stalking
19.6.2 Sexual Harassment or Child Pornography
19.6.3 Forgery
19.6.4 Crime Related to Privacy of Software and Network Resources
19.6.5 Cyber Terrorism
19.6.6 Phishing, Vishing, and Smishing
19.6.7 Malfunction
19.6.8 Server Hacking
19.6.9 Spreading Virus
19.6.10 Spamming, Cross Site Scripting, and Web Jacking
19.7 Various Types of Cyber Attacks in Information Security
19.7.1 Web-Based Attacks in Information Security
19.7.2 System-Based Attacks in Information Security
19.8 Cybersecurity and Privacy Techniques
19.8.1 Authentication and Authorization
19.8.2 Cryptography
19.8.3 Installation of Antivirus
19.8.4 Digital Signature
19.8.5 Firewall
19.8.6 Steganography
19.9 Essential Elements of Cybersecurity
19.10 Basic Security Concerns for Cybersecurity
19.10.1 Precaution
19.10.2 Maintenance
19.10.3 Reactions
19.11 Cybersecurity Layered Stack
19.12 Basic Security and Privacy Check List
19.13 Future Challenges of Cybersecurity
References
20 Changing the Conventional Banking System through Blockchain
20.1 Introduction
20.1.1 Introduction to Blockchain
20.1.2 Classification of Blockchains
20.1.3 Need for Blockchain Technology
20.1.4 Comparison of Blockchain and Cryptocurrency
20.1.5 Types of Consensus Mechanism
20.1.6 Proof of Work
20.1.7 Proof of Stake
20.2 Literature Survey
20.2.1 The History of Blockchain Technology
20.2.2 Early Years of Blockchain Technology: 1991–2008
20.2.3 Literature Review
20.2.4 Analysis
20.3 Methodology and Tools
20.3.1 Methodology
20.3.2 Flow Chart
20.3.3 Tools and Configuration
20.4 Experiment
20.4.1 Steps of Implementation
20.4.2 Screenshots of Experiment
20.5 Results
20.6 Conclusion
20.7 Future Scope
20.7.1 Blockchain as a Service (BaaS) is Gaining Adoption From Enterprises
References
21 A Secured Online Voting System by Using Blockchain as the Medium
21.1 Blockchain-Based Online Voting System
21.1.1 Introduction
21.1.2 Structure of a Block in a Blockchain System
21.1.3 Function of Segments in a Block of the Blockchain
21.1.4 SHA-256 Hashing on the Blockchain
21.1.5 Interaction Involved in Blockchain-Based Online Voting System
21.1.6 Online Voting System Using Blockchain – Framework
21.2 Literature Review
21.2.1 Literature Review Outline
21.2.2 Comparing the Existing Online Voting System
References
22 Artificial Intelligence and Cybersecurity: Current Trends and Future Prospects
22.1 Introduction
22.2 Literature Review
22.3 Different Variants of Cybersecurity in Action
22.4 Importance of Cybersecurity in Action
22.5 Methods for Establishing a Strategy for Cybersecurity
22.6 The Influence of Artificial Intelligence in the Domain of Cybersecurity
22.7 Where AI Is Actually Required to Deal With Cybersecurity
22.8 Challenges for Cybersecurity in Current State of Practice
22.9 Conclusion
References
Index
EULA

Citation preview

The Smart Cyber Ecosystem for Sustainable Development

Scrivener Publishing 100 Cummings Center, Suite 541J Beverly, MA 01915-6106 Publishers at Scrivener Martin Scrivener ([email protected]) Phillip Carmical ([email protected])

The Smart Cyber Ecosystem for Sustainable Development

Edited by

Pardeep Kumar Vishal Jain Vasaki Ponnusamy

This edition first published 2021 by John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA and Scrivener Publishing LLC, 100 Cummings Center, Suite 541J, Beverly, MA 01915, USA © 2021 Scrivener Publishing LLC For more information about Scrivener publications please visit www.scrivenerpublishing.com. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions. Wiley Global Headquarters 111 River Street, Hoboken, NJ 07030, USA For details of our global editorial offices, customer services, and more information about Wiley products visit us at www. wiley.com. Limit of Liability/Disclaimer of Warranty While the publisher and authors have used their best efforts in preparing this work, they make no rep­resentations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchant-­ability or fitness for a particular purpose. No warranty may be created or extended by sales representa­tives, written sales materials, or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further informa­tion does not mean that the publisher and authors endorse the information or services the organiza­tion, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Library of Congress Cataloging-in-Publication Data ISBN 978-1-119-76164-8 Cover image: Pixabay.Com Cover design by Russell Richardson Set in size of 11pt and Minion Pro by Manila Typesetting Company, Makati, Philippines Printed in the USA 10 9 8 7 6 5 4 3 2 1

Contents Preface xxi Part 1: Internet of Things

1

1 Voyage of Internet of Things in the Ocean of Technology Tejaskumar R. Ghadiyali, Bharat C. Patel and Manish M. Kayasth 1.1 Introduction 1.1.1 Characteristics of IoT 1.1.2 IoT Architecture 1.1.3 Merits and Demerits of IoT 1.2 Technological Evolution Toward IoT 1.3 IoT-Associated Technology 1.4 Interoperability in IoT 1.5 Programming Technologies in IoT 1.5.1 Arduino 1.5.2 Raspberry Pi 1.5.3 Python 1.6 IoT Applications Conclusion References

3

2 AI for Wireless Network Optimization: Challenges and Opportunities Murad Abusubaih 2.1 Introduction to AI 2.2 Self-Organizing Networks 2.2.1 Operation Principle of Self-Organizing Networks 2.2.2 Self-Configuration 2.2.3 Self-Optimization 2.2.4 Self-Healing 2.2.5 Key Performance Indicators 2.2.6 SON Functions 2.3 Cognitive Networks 2.4 Introduction to Machine Learning 2.4.1 ML Types 2.4.2 Components of ML Algorithms 2.4.3 How do Machines Learn?

3 4 5 6 7 8 14 15 15 17 18 19 22 22 25 25 27 27 28 28 28 29 29 29 30 31 31 32 v

vi  Contents 2.4.3.1 Supervised Learning 2.4.3.2 Unsupervised Learning 2.4.3.3 Semi-Supervised Learning 2.4.3.4 Reinforcement Learning 2.4.4 ML and Wireless Networks 2.5 Software-Defined Networks 2.5.1 SDN Architecture 2.5.2 The OpenFlow Protocol 2.5.3 SDN and ML 2.6 Cognitive Radio Networks 2.6.1 Sensing Methods 2.7 ML for Wireless Networks: Challenges and Solution Approaches 2.7.1 Cellular Networks 2.7.1.1 Energy Saving 2.7.1.2 Channel Access and Assignment 2.7.1.3 User Association and Load Balancing 2.7.1.4 Traffic Engineering 2.7.1.5 QoS/QoE Prediction 2.7.1.6 Security 2.7.2 Wireless Local Area Networks 2.7.2.1 Access Point Selection 2.7.2.2 Interference Mitigation 2.7.2.3 Channel Allocation and Channel Bonding 2.7.2.4 Latency Estimation and Frame Length Selection 2.7.2.5 Handover 2.7.3 Cognitive Radio Networks References

32 33 35 35 36 36 37 38 39 39 41 41 42 42 42 43 44 45 45 46 47 48 49 49 49 50 50

3 An Overview on Internet of Things (IoT) Segments and Technologies Amarjit Singh 3.1 Introduction 3.2 Features of IoT 3.3 IoT Sensor Devices 3.4 IoT Architecture 3.5 Challenges and Issues in IoT 3.6 Future Opportunities in IoT 3.7 Discussion 3.8 Conclusion References

57

4 The Technological Shift: AI in Big Data and IoT Deepti Sharma, Amandeep Singh and Sanyam Singhal 4.1 Introduction 4.2 Artificial Intelligence 4.2.1 Machine Learning 4.2.2 Further Development in the Domain of Artificial Intelligence

69

57 59 59 61 62 63 64 65 65

69 71 71 73

Contents  vii 4.2.3 Programming Languages for Artificial Intelligence 4.2.4 Outcomes of Artificial Intelligence 4.3 Big Data 4.3.1 Artificial Intelligence Methods for Big Data 4.3.2 Industry Perspective of Big Data 4.3.2.1 In Medical Field 4.3.2.2 In Meteorological Department 4.3.2.3 In Industrial/Corporate Applications and Analytics 4.3.2.4 In Education 4.3.2.5 In Astronomy 4.4 Internet of Things 4.4.1 Interconnection of IoT With AoT 4.4.2 Difference Between IIoT and IoT 4.4.3 Industrial Approach for IoT 4.5 Technical Shift in AI, Big Data, and IoT 4.5.1 Industries Shifting to AI-Enabled Big Data Analytics 4.5.2 Industries Shifting to AI-Powered IoT Devices 4.5.3 Statistical Data of These Shifts 4.6 Conclusion References 5 IoT’s Data Processing Using Spark Ankita Bansal and Aditya Atri 5.1 Introduction 5.2 Introduction to Apache Spark 5.2.1 Advantages of Apache Spark 5.2.2 Apache Spark’s Components 5.3 Apache Hadoop MapReduce 5.3.1 Limitations of MapReduce 5.4 Resilient Distributed Dataset (RDD) 5.4.1 Features and Limitations of RDDs 5.5 DataFrames 5.6 Datasets 5.7 Introduction to Spark SQL 5.7.1 Spark SQL Architecture 5.7.2 Spark SQL Libraries 5.8 SQL Context Class in Spark 5.9 Creating Dataframes 5.9.1 Operations on DataFrames 5.10 Aggregations 5.11 Running SQL Queries on Dataframes 5.12 Integration With RDDs 5.12.1 Inferring the Schema Using Reflection 5.12.2 Specifying the Schema Programmatically 5.13 Data Sources 5.13.1 JSON Datasets

74 74 75 77 77 78 78 79 79 79 80 81 81 82 82 83 84 84 85 86 91 91 92 93 93 94 94 95 95 96 97 98 99 100 100 101 102 103 103 104 104 104 104 105

viii  Contents 5.13.2 Hive Tables 5.13.3 Parquet Files 5.14 Operations on Data Sources 5.15 Industrial Applications 5.16 Conclusion References

105 106 106 107 108 108

6 SE-TEM: Simple and Efficient Trust Evaluation Model for WSNs 111 Tayyab Khan and Karan Singh 6.1 Introduction 111 6.1.1 Components of WSNs 113 6.1.2 Trust 115 6.1.3 Major Contribution 120 6.2 Related Work 121 6.3 Network Topology and Assumptions 122 6.4 Proposed Trust Model 122 6.4.1 CM to CM (Direct) Trust Evaluation Scheme 123 6.4.2 CM to CM Peer Recommendation (Indirect) Trust Estimation (PRx,y(∆t)) 124 6.4.3 CH-to-CH Direct Trust Estimation 125 6.4.4 BS-to-CH Feedback Trust Calculation 125 6.5 Result and Analysis 126 6.5.1 Severity Analysis 126 6.5.2 Malicious Node Detection 127 6.6 Conclusion and Future Work 128 References 128 7 Smart Applications of IoT Pradeep Kamboj, T. Ratha Jeyalakshmi, P. Thillai Arasu, S. Balamurali and A. Murugan 7.1 Introduction 7.2 Background 7.2.1 Enabling Technologies for Building Intelligent Infrastructure 7.3 Smart City 7.3.1 Benefits of a Smart City 7.3.2 Smart City Ecosystem 7.3.3 Challenges in Smart Cities 7.4 Smart Healthcare 7.4.1 Smart Healthcare Applications 7.4.2 Challenges in Healthcare 7.5 Smart Agriculture 7.5.1 Environment Agriculture Controlling 7.5.2 Advantages 7.5.3 Challenges 7.6 Smart Industries 7.6.1 Advantages

131 131 132 132 136 137 137 138 139 140 141 142 143 143 144 145 147

Contents  ix 7.6.2 Challenges 7.7 Future Research Directions 7.8 Conclusions References

148 149 149 149

8 Sensor-Based Irrigation System: Introducing Technology in Agriculture Rohit Rastogi, Krishna Vir Singh, Mihir Rai, Kartik Sachdeva, Tarun Yadav and Harshit Gupta 8.1 Introduction 8.1.1 Technology in Agriculture 8.1.2 Use and Need for Low-Cost Technology in Agriculture 8.2 Proposed System 8.3 Flow Chart 8.4 Use Case 8.5 System Modules 8.5.1 Raspberry Pi 8.5.2 Arduino Uno 8.5.3 DHT 11 Humidity and Temperature Sensor 8.5.4 Soil Moisture Sensor 8.5.5 Solenoid Valve 8.5.6 Drip Irrigation Kit 8.5.7 433 MHz RF Module 8.5.8 Mobile Application 8.5.9 Testing Phase 8.6 Limitations 8.7 Suggestions 8.8 Future Scope 8.9 Conclusion Acknowledgement References Suggested Additional Readings Key Terms and Definitions Appendix Example Code

153

9 Artificial Intelligence: An Imaginary World of Machine Bharat C. Patel, Manish M. Kaysth and Tejaskumar R. Ghadiyali 9.1 The Dawn of Artificial Intelligence 9.2 Introduction 9.3 Components of AI 9.3.1 Machine Reasoning 9.3.2 Natural Language Processing 9.3.3 Automated Planning 9.3.4 Machine Learning 9.4 Types of Artificial Intelligence 9.4.1 Artificial Narrow Intelligence

167

153 154 154 154 157 158 158 158 158 158 160 160 160 160 160 161 162 162 162 163 163 163 164 164 165 166

167 169 170 170 171 171 171 172 172

x  Contents 9.4.2 Artificial General Intelligence 9.4.3 Artificial Super Intelligence 9.5 Application Area of AI 9.6 Challenges in Artificial Intelligence 9.7 Future Trends in Artificial Intelligence 9.8 Practical Implementation of AI Application References 10 Impact of Deep Learning Techniques in IoT M. Chandra Vadhana, P. Shanthi Bala and Immanuel Zion Ramdinthara 10.1 Introduction 10.2 Internet of Things 10.2.1 Characteristics of IoT 10.2.2 Architecture of IoT 10.2.2.1 Smart Device/Sensor Layer 10.2.2.2 Gateways and Networks 10.2.2.3 Management Service Layer 10.2.2.4 Application Layer 10.2.2.5 Interoperability of IoT 10.2.2.6 Security Requirements at a Different Layer of IoT 10.2.2.7 Future Challenges for IoT 10.2.2.8 Privacy and Security 10.2.2.9 Cost and Usability 10.2.2.10 Data Management 10.2.2.11 Energy Preservation 10.2.2.12 Applications of IoT 10.2.2.13 Essential IoT Technologies 10.2.2.14 Enriching the Customer Value 10.2.2.15 Evolution of the Foundational IoT Technologies 10.2.2.16 Technical Challenges in the IoT Environment 10.2.2.17 Security Challenge 10.2.2.18 Chaos Challenge 10.2.2.19 Advantages of IoT 10.2.2.20 Disadvantages of IoT 10.3 Deep Learning 10.3.1 Models of Deep Learning 10.3.1.1 Convolutional Neural Network 10.3.1.2 Recurrent Neural Networks 10.3.1.3 Long Short-Term Memory 10.3.1.4 Autoencoders 10.3.1.5 Variational Autoencoders 10.3.1.6 Generative Adversarial Networks 10.3.1.7 Restricted Boltzmann Machine 10.3.1.8 Deep Belief Network 10.3.1.9 Ladder Networks

173 174 175 176 177 179 182 185 185 186 187 187 187 187 188 188 188 190 190 190 191 191 191 191 193 195 196 196 197 197 198 198 198 199 199 199 200 200 201 201 201 201 202

Contents  xi 10.3.2 Applications of Deep Learning 10.3.2.1 Industrial Robotics 10.3.2.2 E-Commerce Industries 10.3.2.3 Self-Driving Cars 10.3.2.4 Voice-Activated Assistants 10.3.2.5 Automatic Machine Translation 10.3.2.6 Automatic Handwriting Translation 10.3.2.7 Predicting Earthquakes 10.3.2.8 Object Classification in Photographs 10.3.2.9 Automatic Game Playing 10.3.2.10 Adding Sound to Silent Movies 10.3.3 Advantages of Deep Learning 10.3.4 Disadvantages of Deep Learning 10.3.5 Deployment of Deep Learning in IoT 10.3.6 Deep Learning Applications in IoT 10.3.6.1 Image Recognition 10.3.6.2 Speech/Voice Recognition 10.3.6.3 Indoor Localization 10.3.6.4 Physiological and Psychological Detection 10.3.6.5 Security and Privacy 10.3.7 Deep Learning Techniques on IoT Devices 10.3.7.1 Network Compression 10.3.7.2 Approximate Computing 10.3.7.3 Accelerators 10.3.7.4 Tiny Motes 10.4 IoT Challenges on Deep Learning and Future Directions 10.4.1 Lack of IoT Dataset 10.4.2 Pre-Processing 10.4.3 Challenges of 6V’s 10.4.4 Deep Learning Limitations 10.5 Future Directions of Deep Learning 10.5.1 IoT Mobile Data 10.5.2 Integrating Contextual Information 10.5.3 Online Resource Provisioning for IoT Analytics 10.5.4 Semi-Supervised Analytic Framework 10.5.5 Dependable and Reliable IoT Analytics 10.5.6 Self-Organizing Communication Networks 10.5.7 Emerging IoT Applications 10.5.7.1 Unmanned Aerial Vehicles 10.5.7.2 Virtual/Augmented Reality 10.5.7.3 Mobile Robotics 10.6 Common Datasets for Deep Learning in IoT 10.7 Discussion 10.8 Conclusion References

202 202 202 202 202 202 203 203 203 203 203 203 203 203 204 204 204 204 205 205 205 205 206 206 206 206 206 207 207 207 207 207 208 208 208 208 208 208 209 209 209 209 209 211 211

xii  Contents

Part 2: Artificial Intelligence in Healthcare 11 Non-Invasive Process for Analyzing Retinal Blood Vessels Using Deep Learning Techniques Toufique A. Soomro, Ahmed J. Afifi, Pardeep Kumar, Muhammad Usman Keerio, Saleem Ahmed and Ahmed Ali 11.1 Introduction 11.2 Existing Methods Review 11.3 Methodology 11.3.1 Architecture of Stride U-Net 11.3.2 Loss Function 11.4 Databases and Evaluation Metrics 11.4.1 CNN Implementation Details 11.5 Results and Analysis 11.5.1 Evaluation on DRIVE and STARE Databases 11.5.2 Comparative Analysis 11.6 Concluding Remarks References 12 Existing Trends in Mental Health Based on IoT Applications: A Systematic Review Muhammad Ali Nizamani, Muhammad Ali Memon and Pirah Brohi 12.1 Introduction 12.2 Methodology 12.3 IoT in Mental Health 12.4 Mental Healthcare Applications and Services Based on IoT 12.5 Benefits of IoT in Mental Health 12.5.1 Reduction in Treatment Cost 12.5.2 Reduce Human Error 12.5.3 Remove Geographical Barriers 12.5.4 Less Paperwork and Documentation 12.5.5 Early Stage Detection of Chronic Disorders 12.5.6 Improved Drug Management 12.5.7 Speedy Medical Attention 12.5.8 Reliable Results of Treatment 12.6 Challenges in IoT-Based Mental Healthcare Applications 12.6.1 Scalability 12.6.2 Trust 12.6.3 Security and Privacy Issues 12.6.4 Interoperability Issues 12.6.5 Computational Limits 12.6.6 Memory Limitations 12.6.7 Communications Media 12.6.8 Devices Multiplicity 12.6.9 Standardization 12.6.10 IoT-Based Healthcare Platforms

215 217 217 221 223 223 225 225 226 227 227 227 229 230 235 235 237 238 238 241 241 241 241 241 241 242 242 242 242 242 242 243 243 243 243 244 244 244 244

Contents  xiii 12.6.11 Network Type 12.6.12 Quality of Service 12.7 Blockchain in IoT for Healthcare 12.8 Results and Discussion 12.9 Limitations of the Survey 12.10 Conclusion References

244 245 245 246 247 247 247

13 Monitoring Technologies for Precision Health Rehab A. Rayan and Imran Zafar 13.1 Introduction 13.2 Applications of Monitoring Technologies 13.2.1 Everyday Life Activities 13.2.2 Sleeping and Stress 13.2.3 Breathing Patterns and Respiration 13.2.4 Energy and Caloric Consumption 13.2.5 Diabetes, Cardiac, and Cognitive Care 13.2.6 Disability and Rehabilitation 13.2.7 Pregnancy and Post-Procedural Care 13.3 Limitations 13.3.1 Quality of Data and Reliability 13.3.2 Safety, Privacy, and Legal Concerns 13.4 Future Insights 13.4.1 Consolidating Frameworks 13.4.2 Monitoring and Intervention 13.4.3 Research and Development 13.5 Conclusions References

251

14 Impact of Artificial Intelligence in Cardiovascular Disease Mir Khan, Saleem Ahmed, Pardeep Kumar and Dost Muhammad Saqib Bhatti 14.1 Artificial Intelligence 14.2 Machine Learning 14.3 The Application of AI in CVD 14.3.1 Precision Medicine 14.3.2 Clinical Prediction 14.3.3 Cardiac Imaging Analysis 14.4 Future Prospect 14.5 PUAI and Novel Medical Mode 14.5.1 Phenomenon of PUAI 14.5.2 Novel Medical Model 14.6 Traditional Mode 14.6.1 Novel Medical Mode Plus PUAI 14.7 Representative Calculations of AI 14.8 Overview of Pipeline for Image-Based Machine Learning Diagnosis References

261

251 252 253 253 254 254 254 254 255 255 255 256 256 256 256 257 257 257

261 262 263 263 263 264 264 265 265 266 266 266 268 268 270

xiv  Contents 15 Healthcare Transformation With Clinical Big Data Predictive Analytics Muhammad Suleman Memon, Pardeep Kumar, Azeem Ayaz Mirani, Mumtaz Qabulio, Sumera Naz Pathan and Asia Khatoon Soomro 15.1 Introduction 15.1.1 Big Data in Health Sector 15.1.2 Data Structure Produced in Health Sectors 15.2 Big Data Challenges in Healthcare 15.2.1 Big Data in Computational Healthcare 15.2.2 Big Data Predictive Analytics in Healthcare 15.2.3 Big Data for Adapted Healthcare 15.3 Cloud Computing and Big Data in Healthcare 15.4 Big Data Healthcare and IoT 15.5 Wearable Devices for Patient Health Monitoring 15.6 Big Data and Industry 4.0 15.7 Conclusion References

273 273 275 275 276 276 276 277 278 278 282 283 283 284

16 Computing Analysis of Yajna and Mantra Chanting as a Therapy: A Holistic Approach for All by Indian Continent Amidst Pandemic Threats 287 Rohit Rastogi, Mamta Saxena, D.K. Chaturvedi, Mayank Gupta, Mukund Rastogi, Prajwal Srivatava, Mohit Jain, Pradeep Kumar, Ujjawal Sharma, Rohan Choudhary and Neha Gupta 16.1 Introduction 287 16.1.1 The Stats of Different Diseases, Comparative Observation on Symptoms, and Mortality Rate 287 16.1.2 Precautionary Guidelines Followed in Indian Continent 288 16.1.3 Spiritual Guidelines in Indian Society 289 16.1.3.1 Spiritual Defense Against Global Corona by Swami Bhoomananda Tirtha of Trichura, Kerala, India 289 16.1.4 Veda Vigyaan: Ancient Vedic Knowledge 289 16.1.5 Yagyopathy Researches, Say, Smoke of Yagya is Boon 289 16.1.6 The Yagya Samagri 290 16.2 Literature Survey 290 16.2.1 Technical Aspects of Yajna and Mantra Therapy 290 16.2.2 Mantra Chanting and Its Science 290 16.2.3 Yagya Medicine (Yagyopathy) 290 16.2.4 The Medicinal HavanSamagri Components 291 16.2.4.1 Special Havan Ingredients to Fight Against Infectious Diseases 291 16.2.5 Scientific Benefits of Havan 291 16.3 Experimental Setup Protocols With Results 292 16.3.1 Subject Sample Distribution 295 16.3.1.1 Area Wise Distribution 295 16.3.2 Conclusion and Discussion Through Experimental Work 295 16.4 Future Scope and Limitations 297 16.5 Novelty 298

Contents  xv 16.6 Recommendations 16.7 Applications of Yajna Therapy 16.8 Conclusions Acknowledgement References Key Terms and Definitions 17 Extraction of Depression Symptoms From Social Networks Bhavna Chilwal and Amit Kumar Mishra 17.1 Introduction 17.1.1 Diagnosis and Treatments 17.2 Data Mining in Healthcare 17.2.1 Text Mining 17.3 Social Network Sites 17.4 Symptom Extraction Tool 17.4.1 Data Collection 17.4.2 Data Processing 17.4.3 Data Analysis 17.5 Sentiment Analysis 17.5.1 Emotion Analysis 17.5.2 Behavioral Analysis 17.6 Conclusion References

Part 3: Cybersecurity

298 299 299 299 299 304 307 307 309 310 310 311 312 313 313 314 316 318 318 319 320

323

18 Fog Computing Perspective: Technical Trends, Security Practices, and Recommendations 325 C. Kaviyazhiny, P. Shanthi Bala and A.S. Gowri 18.1 Introduction 325 18.2 Characteristics of Fog Computing 326 18.3 Reference Architecture of Fog Computing 328 18.4 CISCO IOx Framework 329 18.5 Security Practices in CISCO IOx 330 18.5.1 Potential Attacks on IoT Architecture 330 18.5.2 Perception Layer (Sensing) 331 18.5.3 Network Layer 331 18.5.4 Service Layer (Support) 332 18.5.5 Application Layer (Interface) 333 18.6 Security Issues in Fog Computing 333 18.6.1 Virtualization Issues 333 18.6.2 Web Security Issues 334 18.6.3 Internal/External Communication Issues 335 18.6.4 Data Security Related Issues 336 18.6.5 Wireless Security Issues 337 18.6.6 Malware Protection 338 18.7 Machine Learning for Secure Fog Computing 338

xvi  Contents 18.7.1 Layer 1 Cloud 18.7.2 Layer 2 Fog Nodes For The Community 18.7.3 Layer 3 Fog Node for Their Neighborhood 18.7.4 Layer 4 Sensors 18.8 Existing Security Solution in Fog Computing 18.8.1 Privacy-Preserving in Fog Computing 18.8.2 Pseudocode for Privacy Preserving in Fog Computing 18.8.3 Pseudocode for Feature Extraction 18.8.4 Pseudocode for Adding Gaussian Noise to the Extracted Feature 18.8.5 Pseudocode for Encrypting Data 18.8.6 Pseudocode for Data Partitioning 18.8.7 Encryption Algorithms in Fog Computing 18.9 Recommendation and Future Enhancement 18.9.1 Data Encryption 18.9.2 Preventing from Cache Attacks 18.9.3 Network Monitoring 18.9.4 Malware Protection 18.9.5 Wireless Security 18.9.6 Secured Vehicular Network 18.9.7 Secure Multi-Tenancy 18.9.8 Backup and Recovery 18.9.9 Security with Performance 18.10 Conclusion References 19 Cybersecurity and Privacy Fundamentals Ravi Verma 19.1 Introduction 19.2 Historical Background and Evolution of Cyber Crime 19.3 Introduction to Cybersecurity 19.3.1 Application Security 19.3.2 Information Security 19.3.3 Recovery From Failure or Disaster 19.3.4 Network Security 19.4 Classification of Cyber Crimes 19.4.1 Internal Attacks 19.4.2 External Attacks 19.4.3 Unstructured Attack 19.4.4 Structured Attack 19.5 Reasons Behind Cyber Crime 19.5.1 Making Money 19.5.2 Gaining Financial Growth and Reputation 19.5.3 Revenge 19.5.4 For Making Fun 19.5.5 To Recognize 19.5.6 Business Analysis and Decision Making

339 340 340 341 341 341 342 343 343 344 344 345 345 345 346 346 347 347 347 348 348 348 349 349 353 353 354 355 356 356 356 357 357 357 358 358 358 358 359 359 359 359 359 359

Contents  xvii 19.6 Various Types of Cyber Crime 19.6.1 Cyber Stalking 19.6.2 Sexual Harassment or Child Pornography 19.6.3 Forgery 19.6.4 Crime Related to Privacy of Software and Network Resources 19.6.5 Cyber Terrorism 19.6.6 Phishing, Vishing, and Smishing 19.6.7 Malfunction 19.6.8 Server Hacking 19.6.9 Spreading Virus 19.6.10 Spamming, Cross Site Scripting, and Web Jacking 19.7 Various Types of Cyber Attacks in Information Security 19.7.1 Web-Based Attacks in Information Security 19.7.2 System-Based Attacks in Information Security 19.8 Cybersecurity and Privacy Techniques 19.8.1 Authentication and Authorization 19.8.2 Cryptography 19.8.2.1 Symmetric Key Encryption 19.8.2.2 Asymmetric Key Encryption 19.8.3 Installation of Antivirus 19.8.4 Digital Signature 19.8.5 Firewall 19.8.6 Steganography 19.9 Essential Elements of Cybersecurity 19.10 Basic Security Concerns for Cybersecurity 19.10.1 Precaution 19.10.2 Maintenance 19.10.3 Reactions 19.11 Cybersecurity Layered Stack 19.12 Basic Security and Privacy Check List 19.13 Future Challenges of Cybersecurity References 20 Changing the Conventional Banking System through Blockchain Khushboo Tripathi, Neha Bhateja and Ashish Dhillon 20.1 Introduction 20.1.1 Introduction to Blockchain 20.1.2 Classification of Blockchains 20.1.2.1 Public Blockchain 20.1.2.2 Private Blockchain 20.1.2.3 Hybrid Blockchain 20.1.2.4 Consortium Blockchain 20.1.3 Need for Blockchain Technology 20.1.3.1 Bitcoin vs. Mastercard Transactions: A Summary 20.1.4 Comparison of Blockchain and Cryptocurrency 20.1.4.1 Distributed Ledger Technology (DLT)

359 360 360 360 360 360 360 361 361 361 361 361 362 364 365 365 366 367 367 367 367 369 369 370 371 372 372 373 373 374 374 376 379 379 379 381 381 382 382 382 383 383 384 384

xviii  Contents 20.1.5 Types of Consensus Mechanism 20.1.5.1 Consensus Algorithm: A Quick Background 20.1.6 Proof of Work 20.1.7 Proof of Stake 20.1.7.1 Delegated Proof of Stake 20.1.7.2 Byzantine Fault Tolerance 20.2 Literature Survey 20.2.1 The History of Blockchain Technology 20.2.2 Early Years of Blockchain Technology: 1991–2008 20.2.2.1 Evolution of Blockchain: Phase 1—Transactions 20.2.2.2 Evolution of Blockchain: Phase 2—Contracts 20.2.2.3 Evolution of Blockchain: Phase 3—Applications 20.2.3 Literature Review 20.2.4 Analysis 20.3 Methodology and Tools 20.3.1 Methodology 20.3.2 Flow Chart 20.3.3 Tools and Configuration 20.4 Experiment 20.4.1 Steps of Implementation 20.4.2 Screenshots of Experiment 20.5 Results 20.6 Conclusion 20.7 Future Scope 20.7.1 Blockchain as a Service (BaaS) is Gaining Adoption From Enterprises References

385 385 386 387 387 388 388 388 389 389 390 390 391 392 392 392 393 394 394 394 397 398 400 401 401 402

21 A Secured Online Voting System by Using Blockchain as the Medium 405 Leslie Mark, Vasaki Ponnusamy, Arya Wicaksana, Basilius Bias Christyono and Moeljono Widjaja 21.1 Blockchain-Based Online Voting System 405 21.1.1 Introduction 405 21.1.2 Structure of a Block in a Blockchain System 406 21.1.3 Function of Segments in a Block of the Blockchain 406 21.1.4 SHA-256 Hashing on the Blockchain 407 21.1.5 Interaction Involved in Blockchain-Based Online Voting System 409 21.1.6 Online Voting System Using Blockchain – Framework 409 21.2 Literature Review 410 21.2.1 Literature Review Outline 410 21.2.1.1 Online Voting System Based on Cryptographic and Stego-Cryptographic Model 410 21.2.1.2 Online Voting System Based on Visual Cryptography 411 21.2.1.3 Online Voting System Using Biometric Security and Steganography 412

Contents  xix 21.2.1.4 Cloud-Based Secured Online Voting System Using Homomorphic Encryption 414 21.2.1.5 An Online Voting System Based on a Secured Blockchain 416 21.2.1.6 Online Voting System Using Fingerprint Biometric and Crypto-Watermarking Approach 417 21.2.1.7 Online Voting System Using Iris Recognition 418 21.2.1.8 Online Voting System Based on NID and SIM 420 21.2.1.9 Online Voting System Using Image Steganography and Visual Cryptography 422 21.2.1.10 Online Voting System Using Secret Sharing–Based Authentication 425 21.2.2 Comparing the Existing Online Voting System 427 References 430 22 Artificial Intelligence and Cybersecurity: Current Trends and Future Prospects Abhinav Juneja, Sapna Juneja, Vikram Bali, Vishal Jain and Hemant Upadhyay 22.1 Introduction 22.2 Literature Review 22.3 Different Variants of Cybersecurity in Action 22.4 Importance of Cybersecurity in Action 22.5 Methods for Establishing a Strategy for Cybersecurity 22.6 The Influence of Artificial Intelligence in the Domain of Cybersecurity 22.7 Where AI Is Actually Required to Deal With Cybersecurity 22.8 Challenges for Cybersecurity in Current State of Practice 22.9 Conclusion References

431 431 432 432 433 434 434 437 438 438 438

Index 443

Preface The cyber ecosystem consists of a huge number of different entities that work and interact with each other in a highly diversified manner. In this era, when the world is surrounded by many unseen challenges and when its population is increasing and resources are decreasing, scientists, researchers, academicians, industrialists, government agencies and other stakeholders are looking forward to smart and intelligent cyber systems that can guarantee sustainable development for a better and healthier ecosystem. The main actors of this cyber ecosystem include the Internet of Things (IoT), artificial intelligence (AI), and the mechanisms providing cybersecurity. This book attempts to collect and publish innovative ideas, emerging trends, implementation experiences, and pertinent use cases for the purpose of serving mankind and societies with sustainable societal development. As outlined in the Table of Contents, the 22 chapters of the book are divided into three parts: Part I deals with the Internet of Things, Part II focuses on artificial intelligence and especially its applications in healthcare, whereas Part III investigates the different cybersecurity mechanisms. In conclusion, we would like to express our great appreciation to all of those with whom we had the pleasure of working with during this project. First, the editors would like to express their deep and sincere gratitude to all the authors who shared their ideas, expertise, and experience and submitted their chapters in a timely manner. Next, the editors wish to acknowledge the extraordinary contributions of the reviewers for their valuable and constructive recommendations that improved the quality, coherence, and content presentation of the chapters. Finally, our heartfelt gratitude goes to our family members and friends for their love, prayers, and concern, allowing us to complete this project on time. Dr. Pardeep Kumar Dr. Vishal Jain Dr. Vasaki Ponnusamy July 2021

xxi

Part 1 INTERNET OF THINGS

1 Voyage of Internet of Things in the Ocean of Technology Tejaskumar R. Ghadiyali1*, Bharat C. Patel2 and Manish M. Kayasth1 Udhna Citizen Commerce College & SPB College of Business Administration & SDHG College of BCA & IT, Surat, Gujarat, India 2 Smt. Tanuben and Dr. Manubhai Trivedi College of Information Science, Surat, Gujarat, India 1

Abstract

In this technological era, the voyage of the Internet of Things (IoT) in the ocean of technology is very interesting, innovative, and beneficial to society. In this voyage, we have to deal with many icebergs in the form of technology such as Machine-to-Machine Communications, Cloud Computing, Machine Learning, Big Data, Distributed Systems, Smart Device, and Security. Blending of such technology with the IoT ultimately promises not only intelligent systems talking to each other but also with human beings in real time in varied domains such as Healthcare, Agriculture, Transport, Corporation services, Manufacturing, and other “Smarter” domains. In this chapter, during the voyage of IoT, we will elaborate Introduction (Basics of IoT, Characteristics, Base Architecture of IoT, and Merits and Demerits), Technological Evolution Toward IoT, Associate Technology in IoT, Interoperability in IoT, Introduction to Programming technology associated with IoT and IoT applications, and A special case study with “Smart Farming: A paradigm shift toward sustainable agriculture” which concludes the chapter. Keywords:  IoT, Internet of Things, Associate Technology with IoT, Interoperability in IoT, Programming in IoT, IoT application, IoT in Agriculture, Smart Farming

1.1 Introduction There are several motivated factors that tell us why the voyage of IoT is important in the ocean of technology. Current internet service basically provides a connection of computers and computing devices, whereas the Internet of Things (IoT) has expanded its scope from computers and computing devices to other things around us. IoT interconnects physical objects around us such as at home it can be communicated with lights, fans, air conditioners, refrigerators, microwave ovens, and other Bluetooth-operated devices, and at the workplace, it can be communicated with internet operated machines. In the recent era, *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (3–24) © 2021 Scrivener Publishing LLC

3

4  The Smart Cyber Ecosystem for Sustainable Development such “Things” connected to the internet have crossed over twenty billion. Such things using embedded electronics are going to connect other things around them depending on the application requirements and thus construct a much bigger Inter-network of Things than that of the current internet of computers and computing devices called the Internet of Things (IoT). To do so, IoT devices have to deal with a challenge of interoperability, that means how such different objects can perform inter-communication with each other. So, this is the integral visualization of the IoT. The other motivated factor in IoT technology implementation is of its low-cost IoT hardware. In IoT, connection of low-cost sensors with cloud platforms gives revolutionary results in this technological era. Using a legitimate merger of these technologies, we can track, analyze, and respond to operational data at a large scale. So, this feature leads toward the end of legacy closed, static, and bounded systems technology and creates a new paradigm of omnipresent connectivity. Such omnipresent connectivity enables communication and exchanges useful information between and with everyday objects around us in order to improve quality of human life. When objects can sense the environment and communicate, they become powerful tools for understanding complexity within it. Such smart objects that can interact with human beings are likely to be interacting more with each other automatically (without human intervention) and updating themselves their daily schedules [1]. Such a phenomenon in the 2000s was heading into a new era of ubiquity, the fact of appearing everywhere internet connectivity is not only serving for Anywhere, Anytime but it also gives the surface of connecting Anything. So, this concept will remove a separation between the real world (physical world) and an imaginary world (internet) resulting that real-world interest should be able to get access to online. In this online access, human beings are very less as internet traffic generators and receivers compared to the things (devices) around us. So, as per the Gartner Research, we can define IoT as, “The Internet of Things (IoT) is the network of physical objects that contain embedded technology to communicate and sense or interact with their internal states or the external environment” [30].

1.1.1

Characteristics of IoT

The applicability and scope of IoT depends on its basic characteristics as given below [2]. Connectivity: Permits network accessibility and compatibility. Compatibility provides the ability to consume and produce data from the network while accessibility is the ability to avail network access. Interconnectivity: This characteristic of IoT says that Anything around us can be Interconnected and thereby communicated with others. Heterogeneity: The devices that are used in IoT are different in nature, hardware platform, and other network-related capabilities. Such devices are known as heterogeneous devices that can work together with each other in various networks. Scalability: In IoT, the amount of devices connected with each other to perform communication is very large, i.e., in millions or even in trillions compared to existing internet. So, in this scenario, it is critical to manage and interpret the generated data that ultimately requires scalable data handling techniques. So,

Voyage of IoT in the Ocean of Technology  5 even when internally connected devices such as sensors and other connecting devices increase, it should not affect network performance. Dynamic: The devices connected with each other in IoT are changing not only their status from connected to disconnected, sleep to wake up but they also update their location as well as speed. Thus, it dynamically updates the number of connected devices. Safety: As we have discussed, millions and trillions of devices are connected in IoT, so safety for data generators and data recipients is must. Such safety design should be made available to data in network, network endpoints, and network itself with sufficient scaling. Having such characteristics, IoT architecture can be classified into three different tiers such as Physical Layer, Network Layer, and Application Layer as follows.

1.1.2 IoT Architecture Different architectures have been proposed by different researchers for IoT, out of which the most popular architecture is three-layer architecture as shown in Figure 1.1. As per its name, this architecture has three layers, namely, i) Perception Layer, ii) Network Layer, and iii) Application Layer. This architecture basically provides the basic idea of the IoT that is further divided into five-layer architecture [2, 5, 20]. The Perception Layer is also known as a physical layer that consists of sensors and actuators. Sensors sense information from the surrounding environment and actuators actuate, i.e., perform some actions based on its sensing from the environment. So, actuators are mechanisms that control the system and accordingly act in the environment. So, basically, this layer provides/acts as input to the IoT architecture. The Network Layer acts as an intermediate layer that provides communication between perception layer and network layer using communication devices such as routers and gateways. The basic task of this layer is to connect smart devices and its related servers. As an intermediary layer, it also transmits and processes the input data of sensing devices received from the perception layer. The Application Layer defines various applications in which IoT is deployed. This layer performs application specific tasks and delivers service accordingly. The specific application may be smart farms, smart homes, smart cities, etc. Application Layer

Cloud/Servers

Network Layer

Routers and Gateways

Perception Layer

Sensors

Figure 1.1  Three-layer IoT architecture.

and

Actuators

6  The Smart Cyber Ecosystem for Sustainable Development Furthermore, this three-layer architecture converts into five-layer architecture. In this architecture, the perception layer and application remains and acts the same but the intermediate network layer is again divided into three sub layers such as Transport Layer, Processing Layer, and Business Layer. In this five-layer architecture, the transport layer transmits sensor data from perception layer to processing layer. Processing layer acts as a middleware layer and it can store, analyze, and process data that was received from the transport layer. To do such a task on data, this layer basically deals with many other data management technologies such as Big Data computing to process, cloud computing to store, and database management system to manage the data. The Business Layer acts on the top of the application layer and manages the entire IoT system. It consists of all application-related information such as profit model of business to that application, user privacy to the specific application, and other such business application-related information. Based on this architecture, the merits and demerits of IoT are as follows.

1.1.3 Merits and Demerits of IoT Merits and Demerits of the IoT are as mentioned below: Merits: ✓✓ Access Information: Due to association of cloud computing technology with IoT, the data access of the IoT is easy and available anywhere, anytime. So, a person involved in IoT can have easy access to such information and thereby without having his/her physical presence, it makes it convenient for that person to go about their work. ✓✓ Communication: IoT has important characteristics such as Interoperability. Using such characteristics, devices involved in IoT can easily talk to each other. In this way, inter-device communication will be more transparent and thereby increase efficiency of connected devices, e.g., a production unit comprises IoT technology, has Machine-to-Machine (M2M) communication that makes the product better by reducing inefficiencies, and produces faster results. ✓✓ Cost Effective: This merit is actually a combined benefit of the above two merits. Due to easy information access and communication, IoT devices can transmit the data very quickly over a connected component in the network of IoT. So, it saves time compared to traditional data transfer in which it occupied much more time and in that way IoT became a cost effective solution. ✓✓ Automation: It means to manage routine work without any human intervention. IoT helps in business automation and improves the quality product or service. IoT can collect data from the network and perform analytics on it to reveal business insights and opportunities, and thus reduce operational cost. In the automation process, IoT can also predict needs before they arise and take action accordingly to gain business profit. Demerits: • Privacy and Security: IoT is an inter-network of things that consists of multiple devices interconnected to each other. So, such interconnection might

Voyage of IoT in the Ocean of Technology  7 increase the risk of any leakage of important data. In this scenario confidential information may not be safe and could be fetched/hacked by someone else easily. • Complexity: IoT is not only a collection of interconnected heterogeneous devices but also a combination of heterogeneous networks. So, in this environment, a single ambiguity can affect the entire system tremendously. This certainly creates a complicated state of affairs and easily increases complexity. • Lesser Jobs: Automation is one of the merits of IoT, so the need for manual processing that can be done by human beings will reduce drastically. So, the future of IoT may be one of the reasons for unemployment. • Dependability: Being complex is one of the demerits of IoT. Due to its intra heterogeneous objects and inter-heterogeneous network connectivity, it also increases dependability of such intra/interconnected object(s)/network(s). So, in case of a bug in the system, there may be a change or collapse of the entire system. Day-by-day IoT technology dominates human lifestyle and thereby increases dependability on IoT technology altogether.

1.2 Technological Evolution Toward IoT After passing several decades of invention of an electronic device computer, in the 1960s, a communication between two computers was made possible using a computer network. Functioning of the internet commenced after the invention of TCP/IP in the 1980s. Later on, in 1991, the internet became more popular using available WWW. After the invention of www, e-mail, information sharing, and entertainment were introduced on the internet. Interconnectivity of different objects (devices) evolved over the years, and it became the base for technological evolution toward IoT. Web applications became prevalent with evolved network technology resulting in an internetworked ATM. E-commerce was also introduced during this time. Till 2000, Information and Communication Technology (ICT) provided service related to “anytime”, “anywhere” paradigms. It means it provides service connectivity through the internet any time at any place. But in 2000, we witnessed a new era of ubiquity that suggests a new paradigm of connecting “anything” IoT [20]. Mobile internet technology was also formed parallel to evolution on IoT from 2000 to 2010. Due to the invention of mobile internet technology, social networking platforms such as Skype (2003), Facebook (2004), Twitter (2006), and WhatsApp (2009) were also introduced and thereby the users are getting connected via the internet through connecting devices [3, 4]. As shown in Figure 1.2, IoT technology was infant in 2000, and it has matured during the decade that dealt with other pioneering technologies such as RFID, WSN, and M2M communication that underwent revolution in the product automation industries and service industries. After having M2M communication, IoT which is a network of objects, that communicates with each other via different technologies such as Internet, RFID, GPRS, computers, actuators, and mobile phones without or minimal intervention of human beings. The voyage of IoT technology has been continuing in the path of IoT application domain such as Digital Locker, Smart Healthcare, Smart Vehicle, and Smart Cities. Recently, IoT

8  The Smart Cyber Ecosystem for Sustainable Development THINGS

IoT

INTERNET OF

PEOPLE

Social Media

SERVICE CONTENT PreInternet Before 1980

Web 2.0 WWW

Skype Facebook Twitter Whats App Web Application E-Commerce

Intelligent Device Machine to Machine WSN RFID

E-Mail Information Entainment

Human to Human 1980

....

....

20002010

20102020

2030 ?

Figure 1.2  Evolution of IoT.

technology emphasizes Smart Dust (a smaller computer than a grain of sand) collaboration with evolved nanotechnology to diagnose problems in the object system or human society. Thus, the IoT is a paradigm shift in the Internet technology that is rapidly developing by the advancements in other enabling technologies such as sensor networks, mobile devices, wireless communications, networking, and cloud technologies that results into Industrial IoT (IIoT), an application of IoT in industries. So, now, we will discuss those associated technologies which bring such technological revolution in association with IoT.

1.3 IoT-Associated Technology In Inter-networked of Things, stake holding technologies expect trillions of Sensors, billions of Smart Systems, and millions of applications in near future. There are numerous supportive technologies with IoT to perform smarter than before. IoT Associative Technology can be classified into four sub-topics, namely, (i) (ii) (iii) (iv)

Sensor and Actuators, IoT Networking, IoT Connectivity Technologies, and IoT Communication Protocol.

(i) Sensor and Actuators: Sensor and Actuators are the most essential and core components of the IoT. As per oxford dictionary, the meaning of sensor is “It is a device which detects or measures a physical property and records, indicates or otherwise responds to it.” [31]. So, sensors basically sense the physical observable fact around us from an environment. As per the other sources [32], sensors can be defined as “A sensor detects (senses) changes in the ambient conditions or in the state of another device or a system, and forwards or passes this information in a certain manner”. According to this definition, a sensor can sense or detect the physical phenomena or measured properties such as temperature, humidity, smoke detection, and

Voyage of IoT in the Ocean of Technology  9 obstacle detection. So, we have different specific sensors that can be used to sense particular properties and cannot be used to sense or to detect, or be insensitive to the other properties surrounding us, i.e., specific physical properties can only be detected by specific sensors not bothering about other properties surrounding us. For example, a temperature sensor can sense heat (temperature) around us and then these sensed values are converted into its equivalent electrical signals. The smallest change that can be detected and can be measured by a sensor as an output is known as resolution. Based on the output, the sensor can be classified into two categories: Digital Sensor and Analog Sensor. Analog Sensor can generate or produce a continuous output signal equivalent to continuous measured property in nature; e.g., temperature, humidity, pressure, and speed are analog quantities. While Digital Sensor produces binary output (0 or 1, ON or OFF) signal. So, it generates a non-continuous (discrete) value in the form of bits that combine to gather generated byte as an output. Based on the output data types, sensors can be classified into two major groups: Scalar Sensor and Vector Sensor. Scalar Sensor generates output proportional to quantity measured from surroundings without considering its orientation or direction, e.g., physical quantity such as temperature and pressure. Vector Sensor generates the output that is proportional to quantity measured as well as its orientation or direction, e.g., physical quantities such as sound and velocity. Based on sensed information from sensors, actuators basically perform some actions (actuates) on the physical environment. So, here, actuators take actions based on what has been sensed and in that way controls a system that can be acted upon an environment. In this context, the actuators require some control signal and source of energy to function further. So, when actuators receive such control signals, they convert the energy into mechanical motion. Based on their functional domains, we have three broad categories of actuators such as pressure-based actuators (hydraulic and pneumatic), electric-based actuators (electrical, thermal, and magnetic), and mechanical-based actuators. Other than these types of actuators, other popular actuators are used in industries. Agriculture uses Soft actuators. Soft actuators are polymer-based actuators designed to handle delicate objects and used in robotics. “Transducer” is another associative term which can be used for both Sensors and Actuators [33]. So, actuators sense the surroundings in the form of information and are converted into electrical signals; such control signals are received by actuators and action is taken accordingly. For example, in “soil moisture and water level monitoring application”, agriculture soil water/moisture level in a farm is sensed by specific sensors, is converted into electrical signal, and is provided to the actuator as “solenoid valve”. Solenoid valves consist of a mechanism that allows or stops the water flow. So, depending on the electrical signal received from the sensor (water/moisture level), this solenoid valve as an actuator can actuate, i.e., flow water or stop water.

10  The Smart Cyber Ecosystem for Sustainable Development (ii) IoT Networking: IoT Network consists of several components such as Device (The Thing), Local Network, Internet, Backend services, and Applications. Here, in case of “Device”, it consists of a collection of sensors and actuators that can act as one component in the entire IoT Network. These become different nodes in the IoT Network that can be communicated with each other. As shown in Figure 1.3, a node in IoT Network can be communicated with other target node via another component of IoT network, i.e., Local Network. If target node does not belong to the local network, IoT network will search it through another component of IoT network, i.e., Internet. In Backend services, the data may be received from local networks or from the internet and perform complex analysis using different machine learning algorithms. Such result generated after complex analysis is given to applications that serve as an output of IoT Network. Thus, IoT is a very complex system that involves things (sensors and actuators), local area network, wide area network (internet), machine learning, and analysis algorithms which act mutually into one system entity. Such result generated after complex analysis is given to applications that serve as an output of IoT Network. Thus, IoT is a very complex system that involves things (sensors and actuators), local area network, wide area network (internet), machine learning, and algorithms which act mutually into one system entity. So, to perform function through IoT we need more associative technology such as Bigdata, M2M communication, cloud computing, Cyber Physical System (CPS), 3G/4G/5G, and Internet of Vehicles (IoV). To perform suitable communication among such heterogeneous technologies and devices, we need to deal with certain challenges of IoT. They are securities, interoperability, scalability, energy efficiency, and interfacing. IoT connectivity technologies are involved in IoT communication to execute it properly. They are as per the sub topics given below. (iii) IoT Connectivity Technology: Connectivity among devices is fundamental when we think about the IoT. There are several IoT connectivity technologies in the form of communication protocols that utilize IoT networks to perform communication between IoT devices (Things). IoT service offering protocols such as RFID (Radio Frequency Identification), CoAP (Constrained Application protocol), XMPP (Extensible Messaging

Local Network Device (The Thing)

Backend Services Internet

Figure 1.3  IoT networking.

Applications

Voyage of IoT in the Ocean of Technology  11 Presence Protocol), MQTT (Message Queuing Telemetry Transport), AMQP (Advanced Message Queuing Protocol), and 6LoPAN (IPv6  over Low-Power Wireless Personal Area Networks) are basically utilized to establish connectivity between IoT devices in IoT network. RFID stands for Radio Frequency Identification that is used widely in shopping malls as a system whole. RFID system consists of RFID tag, RFID reader, and RFID software. RFID tag is covered by a hard jacket that consists of integrated circuit and antenna and stores digitally encoded data. RFID tags are categorized into Active Tag (own power supply) and Passive Tag (dependent for power supply). RFID reader reads from the tag and transfers data to RFID software for further processes to operate. CoAP, as per its name Constrained Application Protocol [13], is utilized for web transfer just as http but in constrained networks resources environment such as limited computational resources, limited bandwidth and limited power supply in IoT. CoAP in IoT network functions as a session layer and an application layer. CoAP is designed for M2M communication and uses a request-response model for two connected endpoints (objects) in the IoT network. XMPP stands for Extensible Messaging and Presence Protocol [06], an open standard XML (extensible markup language)–based middleware protocol that is used for real-time structured data exchange. XMPP uses decentralized client-server architecture which means the central server is not located for message transfer. So, in this context, XMPP provides flexibility in sustaining interoperability between different things (objects), between diverse systems, and between heterogeneous protocols in the IoT network. XMPP does not support text-based communication. MQTT is a Message Queuing Telemetry Transport protocol [7], publish-subscribe–based ISO standard protocol. So, in this protocol, publisher publishes the data that can be utilized by the subscriber and this phenomenon creates this protocol as a lightweight protocol that can be used in combination with TCP/IP protocol. In MQTT, there is a central entity known as “broker” which is responsible for transferring messages from sender to receiver. Here, client publishes a message to the broker including topic (routing information for broker). Based on the matching topic, the broker delivers the message to the client. So, in this architecture, client does not know the real message passing entity, this feature provides a highly scalable solution independent of data producer and data consumer. MQTT is used by Microsoft Azure, Amazon Web Services, Facebook Messenger, and Adafruit for providing various services. AMQP stands for Advanced Message Queuing Protocol [14] and is ISO/ IEC-based open standard protocol for passing business messages between different business applications or organizations. At the time of passing business messages, AMQP is persistent and provides three different types of message delivery guarantees. They are At-most-once (message delivered

12  The Smart Cyber Ecosystem for Sustainable Development once or never), At-least-once (message certainly delivered may be multiple times), and Exactly-once (certainly delivered and only once). 6LoPAN is an IPv6 over Low-Power Wireless Personal Area Networks [15]. Due to large components involved in the IoT network, unique address identification can be done through IPv6 (64 bits) address protocol instead of IPv4 (16 bits) address protocol. This protocol provides transmission of data wirelessly with limited data processing potential in PAN. So, as per its name, it permits low-powered devices to connect to the internet which is also a basic characteristic of IoT networks. (iv) IoT Communication Protocol: Other well-known communication protocols that require to perform communication in IoT network are IEEE 802.15.4, Zigbee, Z-Wave, Wireless-HART (wireless  sensor networking technology based on the Highway Addressable Remote Transducer Protocol), Near-Field Communication (NFC), and Bluetooth as given below. IEEE 802.15.4 is an extensively used standard protocol for establishing communication in the IoT network [9, 10]. It provides a framework for lower layers such as physical layer and Mac layer to a small range of Personal Area Network (PAN) and Wireless Personal Area Network (WPAN) that generally range from 10 to 75 meters in the environment of low-power, low-speed, and low-cost requirements. It uses star and peer-to-peer network topologies for establishing communication between neighboring devices in the IoT network. Zigbee is an enhanced version of IEEE 802.15.4 that functions on top of layer 1 and layer 2 of IEEE 802.15.4 in layer 3 and onwards [11, 12]. So, Zigbee uses the MAC layer to the application layer in the IoT Network. Zigbee is basically used for Wireless Sensor Network (WSN) and supports stat and mesh topology. In Ad-hoc network, Zigbee utilizes Ad-hoc On-demand Distance Vector (AODV) Protocol for broadcasting a route request to all its immediate neighbors. Such neighbors spread this message to their neighbors and, in that way, messages can be spread all the way through the IoT network. One of the important applications utilized by Zigbee is “Building Automation”. Other applications are healthcare monitoring, home energy monitoring, LED lighting monitoring, telecom services, and many more. Z-Wave is a well-known protocol for home automation to do different functions using various IoT devices. It functions on mesh topology that can have up to 232 nodes (devices) in a network and uses radio frequency for communication, i.e., signaling and controlling of home automation IoT devices. In a home, there is a Z-Wave controller that controls the signal communication with existing other Z-Wave nodes (devices). Such Z-Wave devices may communicate directly with each other or they can communicate via Z-Wave controller in smart home automation systems. Wireless-HART is a wireless sensor networking technology based on the Highway Addressable Remote Transducer Protocol [16] that is developed

Voyage of IoT in the Ocean of Technology  13 for networked smart field devices. IoT implementation and cost of performing communication between IoT devices will be cheaper and easier using HART. There are certain differences between physical HART and wireless HART in the context of physical layer, data link layer, and network layer. HART physical layer utilizes IEEE 802.15.4 protocol. HART data link layer has a provision of a super frame that ensures suitable communication between different IoT devices (nodes) of the IoT network. Wireless HART network layer uses mesh topology for communication in IoT networking. Wireless HART protocol network layer can be composed of OSI Network layer, transport layer, and session layer. HART application layer is responsible for generating responses by extracting commands from messages and executing them. So, the basic difference between Wireless HART and Zigbee is that Zigbee hops when the entire network hops but Wireless HART hops after every message. Near-Field Communication (NFC) is designed for use of devices in its close proximity and uses magnetic induction principle just as RFID [17]. Based on power/energy resource availability, NFC has two types, viz., Active NFC and Passive NFC. Active NFC does not depend on external power/energy resources and Passive NFC depends on external power/ energy resources. Like RFID, NFC also has three components: reader, tag, and software. NFC reader creates magnetic fields using electric current that connects the physical space between these two devices, NFC reader and NFC tag, and can transmit encoded information from NFC tags such as identification number. NFC can be operated in three different modes much as peer-to-peer, read/write and card emulation. For example, in peer-topeer mode, two smartphones can communicate with each other. In read/ write mode, one active and one passive device is involved to perform communication and in card emulation NFC can be used for contactless credit card operation. Bluetooth is a wireless short range communication technology that is heavily used in establishing communication in IoT network devices in PANs [18]. Bluetooth can be utilized to perform communication between two smart phones for transferring data to short range. Bluetooth uses ad-hoc technology known as Ad-hoc Piconets. Connection establishment in Bluetooth can be possible in sequence using three different phases such as Inquiry, Paging, and Connection. In the “Inquiry” phase, Bluetooth devices discover other Bluetooth devices near it. After finding a Bluetooth device nearer to the current device, in this second phase “paging”, connection can be established between these two devices. In the third phase of “connection”, either devices can actively participate in the network or enter into low-power sleep mode. After discussing IoT associative technology such Sensor and Actuators, IoT Networking, IoT connecting technology, and IoT communication protocol, the important characteristic of IoT devices and technologies that make all this possible is “Interoperability in IoT” as discussed in the next topic.

14  The Smart Cyber Ecosystem for Sustainable Development

1.4 Interoperability in IoT In IoT, many heterogeneous devices, protocols, operating systems have to work together to fulfill objectives. This heterogeneity is one of the major concerns when we perform communication in the world of IoT as it requires not only anytime, anywhere but also anything enabled to communicate. “Interoperability” is a characteristic of a product or system whose interfaces are completely understood to work with other products or systems without any limitations. Interoperability is must when we would like to communicate in the era of IoT that contains heterogeneous devices [19]. So, by maintaining interoperability in the IoT network, we can have exchange of data and service in a seamless manner. In this seamless exchange of data and service, many elements are involved and perform the communication such as physical objects can communicate with other physical objects. As per the overall goal of IoT, anytime anywhere anything (device) can be communicated with other devices, i.e., can do Device-to-Device (D2D) communication. More than these types of communication, others such as Device-to-Machine (D2M) communication, M2M communication should also be performed seamlessly in the IoT network. Hence, in this situation, the IoT network has to deal with many types of heterogeneity such as heterogeneity of different wired and wireless communication protocols. Moreover, different programming languages are used for different platforms as well as different hardwares that also vary different standards and support different languages and communication protocols. So, if we would like to perform seamless communication between such corel, heterogeneous connected components, protocols, languages, operating systems, databases, and hardwares, then interoperability among them is a must. There are basically two types of Interoperability such as User Interoperability and Device Interoperability. User Interoperability is an interoperability problem between user and device(s) and Device Interoperability is an interoperability problem between two different devices. User interoperability problems occur when remotely located users would like to communicate with other device(s) whose product id may be written in different language, there may be differences in user syntaxes, differences in user semantics, as well as differences in user specification for those devices. So, all these types of complex veracity leads to create a simple IoT problem into a complex one that falls under the problem of interoperability. To resolve such user syntax interoperability problems worldwide, there are different solutions that provide unique device identification addresses to devices such as Electronic Product Codes (EPC), Universal Product Code (UPC), Uniform Resource Locator (URL), and IP addresses IPv6. For resolving syntactic interoperability problems there are different approaches such as Open standard protocol (IEEE 802.15.4, IEEE 802.15.1, and Wireless HART), Closed standard protocol (Z-Wave), Service Oriented Computing (SOC), and web services. But all these approaches have the problem of heterogeneity and, therefore, incompatible with each other to perform communication. So, we have certain middleware technologies such as Universal Middleware Bridge (UMB) that resolve such devices interoperability problems that have been generated due to heterogeneity amongst them. Thus, in this topic we have discussed IoT which is surrounded by heterogeneity problems, which can be resolved using interoperability features. The next topic explains about the programming technologies concerned with IoT.

Voyage of IoT in the Ocean of Technology  15

1.5 Programming Technologies in IoT The programming technologies associated with IoT such as Arduino programming, Python programming, and Raspberry Pi are well known. Arduino programming can be done in consultation with the Arduino UNO board to accept analog and digital signals as input and generate desired output. Python is a lightweight programming language that is very much popular for IoT application development. Raspberry Pi is powerful compared to Arduino in terms of memory capacity and processing power. Raspberry Pi is a single-board, low-cost computer that provides easy access using GUI.

1.5.1 Arduino Arduino is low resource consuming and cheaper in cost. Due to these two characteristics, it is popular worldwide for implementing the IoT. As shown in Figure 1.4, Arduino is an open source programmable board with a built-in microcontroller and the software (IDE). So, using this Arduino board, we can have input as analog or digital signals and produce digital signal as an output and there is no need to have a separate programmer to program it like traditional microprocessor 8051 and 8085. To program the Arduino microcontroller board, open source software of Arduino IDE is utilized using C or C++ programming language. IDE can be downloaded from Arduino’s official website [22]. To do programming in the Arduino board, install Arduino IDE. Now switch on the Arduino board by attaching it with USB cable to PC and launch Arduino IDE. Using the TOOLS option of this IDE, set BOARD type and PORT type. Program coded in Arduino is known as “sketch”. So, go to the file menu and click on “Create New Sketch” to write a new program in Arduino. Sketch structure in the Arduino IDE can be divided in two major functions: Setup() and Loop(). Setup() function is just like the main() of C/C++ in which we can declare input/output variables and pinmodes can also be declared over here. As per the name, Loop() function is used for iterating the instruction(s) written under it. Using the

Figure 1.4  Arduino UNO board [21].

16  The Smart Cyber Ecosystem for Sustainable Development “pinMode()” function of the Arduino IDE library, we can have the syntax of this function as given below. The common functions of the Arduino library are as given in Table 1.1. Table 1.1  Arduino function and its description. Function

Function Description

pinMode(pin, Mode)

Configure the input/output pin(s) with its pin number in the arduino board. pin = pin number on Arduino board, Mode = INPUT/ OUTPUT

digitalWriter()

Write digital pin value (HIGH/LOW)

analogRead()

Read from analog input pin

Delay()

Provides a delay of specified time in milliseconds

Using the above common function, we can write down a program in Arduino IDE that is used for “Blinking LED”. To perform this practical on an Arduino board, we require several objects/entities as hardware such as an Arduino micro-controller board, USB connector, LED, respective capacity of resistor, bread board, connecting wires, and as a software Arduino IDE as shown in Table 1.2. Table 1.2  Arduino programming requirement. Activity

Hardware Side

Software Side

Prerequisite

Arduino micro-controller board, USB connector, LED, respective capacity of Registers, Bread board, Connecting wires

Arduino IDE

Process

✓ Using connecting wires, set LED on breadboard and connect it to Arduino Using USB connector, connects Arduino board to PC

• Select Board and Port type • Write equivalent Sketch in Arduino IDE Verify sketch and upload it

Sample Arduino “Sketch” for Blinking LED: void Setup( ) { pinMode(12, OUTPUT); // set arduino pin number 12 for digital output } void Loop( ) { DigitalWriter(12,HIGH); // Turn ON the LED Delay(500); // wait for 500 millisecond = 0.5 second DigitalWriter(12,LOW); //Turn OFF the LED Delay(500); // wait for 500 millisecond = 0.5 second }

Voyage of IoT in the Ocean of Technology  17

1.5.2 Raspberry Pi As shown in Figure 1.5 Raspberry Pi is a low-cost, single-board, palm-size computer that provides easy access. Raspberry Pi has higher processing capabilities and more features compared to Arduino [23]. So, such programming technology is better when we have more data for processing such as image and multimedia sensor data processing. To do so, we can download freely available Raspberry Pi–based operating systems which are GUI-based systems. For example, Raspbian and Noobs are officially supported OS for Raspberry Pi. Other operating systems that also support this technology are Windows 10 core, Snappy Ubuntu code, Ubuntu Mate, Pinet, and Risc OS. Supported programming languages for Raspberry Pi are C, C++, JAVA, Python, and Ruby. The following Raspberry device can act as a server as well as a node in IoT networking. So, we can create an interactive environment using such a network of connected devices. We can have an IoT-based system that can perform different tasks such as collecting data from connected sensors of the network, send such received data to a remote machine or server, process the data, and respond accordingly in the IoT network. For example, suppose we have a digital DHT (Digital Humidity Temperature) sensor that senses the data of the surrounding environment. Collected data is then transferred to server and saved on server for further processing and after processing such information is updated on screen based on responses available from the network. To do so, we require a digital humidity temperature sensor, register, jumper wires, and Raspberry Pi unit. As shown in Figure 1.6, DHT sensors have four pins numbered as 1, 2, 3, and 4. Pin 1 used for power supply of 3.3 to 5.0 V, pin 2 used for data, pin 3 is null, and pin 4 utilized for ground. So, connect pin 1 of DHT sensor to the 3.3V pin of Raspberry Pi, connect pin 2 of DHT sensor to any input pin of Raspberry Pi and connect pin 4 of DHT sensor to ground of Raspberry Pi. So, after establishing connection of DHT sensor with Raspberry Pi, reading data from sensor using “.read_retry” method consists in “Adafruit” library of DHT22 sensor. To transfer data to the server, we can establish a connection between client and server, send data from client to server and then save the data in a particular file at server

Figure 1.5  Raspberry Pi 4 [24].

18  The Smart Cyber Ecosystem for Sustainable Development

Ground(3) Data(2) Vcc(1)

Vcc(1)

Data(2)

Grid(4)

Figure 1.6  DHT sensor [29].

end in the form of a log file. Data processing is done at server end that may include filtering and plotting of data. Due to lack of data, there may be a chance of incomplete or corrupted data so to overcome such data, and we need preprocessing activity such as cleansing and to do so we use filtering over here. To plot the 2D data at the server end, a Python library MATPLOTLIB can be utilized. In this way, using a DHT sensor with Raspberry Pi, we can monitor the value of humidity and temperature on screen with the help of GUI. Even we can extend/update the script and instruct the rotary motor as an actuator to actuate (start fan) when certain room temperature increases. So, these are some of the modest applications of IoT in real-world environments that can be implemented using Raspberry Pi.

1.5.3 Python As a lightweight versatile scripting programming language, Python is very much popular and useful in IoT-based application development [25]. It provides some kind of relaxed environment, i.e., it does not follow strict rules. Python-Integrated Development Environment (IDE) provides several modules and libraries, using which one can establish connectivity with many hardware and also compatible with multiple OS such as windows, Linux, and MAC. Well-known Python IDEs are Spyder and PyCharm. To perform file operation in Python, we do not require any separate library, and it is an in-built function such as open(), read(), write(), and close(). Python supports various file formats to perform such operations like .Txt file and .CSV file. This feature makes data management easier using python programming language. If our data file is of the type of image, then we have a Python Image Library (PIL) to do the process with such a file. In this library, there are famous functions/methods like open(), show(), resize(), rotate(), print(), and convert() to do various tasks on such images that are contained in an image file. Python also supports client-server architecture model and provides necessary network services to it. Socket programming in Python allows us to implement clients and servers for connection oriented as well as connection less protocols. In socket programming of

Voyage of IoT in the Ocean of Technology  19 Python, we have to import “socket” and “sys” libraries that contain well-known and most utilized functions for example connect(), send(), and listen() using which one can establish connection between clients (IoT nodes) and server. Python has also a separate library that provides and deals with a specific application level network protocol(s). Thus, Python is a versatile object oriented programming language that provides an easy environment in open source community software for the development of IoT-based applications.

1.6 IoT Applications With the collaboration and co-operation of other technology involved with IoT, it has vast scope in various IoT-based applications such as Smart Home, Smart Healthcare, Smart Transportation, Smart Asset Management, and Smart Farm [26]. Such applications will create a paradigm shift in the traditional lifestyle of human beings and that is why nowadays the popularity of IoT is much more than other existing technologies. Some of the well-known IoT applications are as given below. Smart Home Smart home as an IoT application contains features like integration of various IoT-enabled devices, provides securities amongst them, and enables networking using central controlled devices and its related security features that adapt a traditional home into technically enriched sophisticated home. Such IoT-enabled devices monitor some important aspects for home such as remote air conditioning, heating, and ventilation management using smart phones. It also performs the operation management by communicating with different IoT-enabled devices of home like IoT-enabled fan, tube light, oven, and washing machine. Smart Healthcare Such smart healthcare applications are also known as the Internet of Medical Things (IoMT). Its popular applications are “Remote Health Monitoring” and “Emergency Notifications System” [27]. There are many devices that can monitor the number of health parameters of human beings. IoT-enabled devices, by collaborating communication with medical manifesto, can monitor the heartbeat and blood pressure and, with proper medical surgery, can also act as pacemaker. “Smart Bed” is an instrumental bed which maintains a patient’s regular checkups without any human intervention (nurse). Moreover, such smart beds can also be connected using smart sensors that can acquire information from the patient end and analyze and transmit them to smart home objects connected to this system. To monitor the well-being of senior citizens, smart sensors can also be medically equipped within living spaces of human beings. Smart Transportation In different aspects of the transportation system, IoT is helpful in doing things more smartly than done earlier. IoT-enabled devices can be equipped with vehicles, infrastructures, drivers and other human beings involved in transporting activities and can play the role of a monitor or supervisor. So, logistics, smart traffic control, vehicle control, and fleet

20  The Smart Cyber Ecosystem for Sustainable Development management are several well-known applications of the Smart Transportation segment. During transportation of any goods container, it can be handled by monitoring the realtime location of the container, the status of the container (open/close), and how the container can be handled throughout the journey. So, such smart tracking can provide security features to that container and thereby minimize the theft risk and maximize the possibilities of recovering stolen material. Smart Asset Management Asset management is one the oldest problems faced by many industries. Asset is basically an instrument or a device that may be cheap or priceless, that may be located indoor or outdoor. So, in case of an emergency, it is often a problem finding/tracking its location in the organization. IoT can provide solutions toward pinpointing the asset’s exact location within a short span of time. For example, in hospitals, there are many assets such as medical instruments, scanning machines, and healthcare monitors loosely coupled with each other. So, by using IoT-enabled solutions, one can correlate them technically and upload the data on cloud to monitor its future activities such as scheduled maintenance without intervention of human beings. There are many other domains too in which IoT can be applied to operate things better and smarter such as Smart Retailing, Smart Inventory Management, Smart Tracking, and Smart Cargo Management. In industries, the IIoT can be applied. That is one of the reasons for Industrial Revolution 4.0. So, in the context of industries IoT, we have other broad domains in which IoT can be served. Such domains are Smart Factory, Food Industries, Plant securities and safety, Oil Chemical and Pharmaceutical Industries, Unmanned Auto Vehicle industries (UAVs), and many more. The domain of agriculture also utilizes IoT facilities in different sub-applications and converts the agriculture farm into a Smart Farm. So, in the next sub-topic, we shall discuss how premium facilities can be developed in traditional farms and how one can use IoT technology to convert a farm into a Smart component of sustainable agriculture. Smart Farm—A Paradigm Shift in Sustainable Agriculture Smart Farm is an IoT application that gives leverage to the farmer community to do many farm level tasks using IoT without human intervention or minimal human intervention. Smart Farm consists of a variety of functions such as water level management, soil fertility management, pesticides control, and many more. IoT-enabled devices can be useful to fulfill the basic communication functionality that result into performing smart work in the agriculture domain at farm level. In future, smart farms can have the facilities such as soil moisture and water level monitoring, automated irrigation system, automated sowing and weeding system, automated organic waste management system, automated environment monitoring system, and soil micronutrients monitoring system as shown in Figure 1.7. ✓✓ Out of these systems, IIT Kharagpur, India, developed an automated irrigation system, “AgriSens” that focused on Smart Water Management using IoT [29]. AgriSens provides automatic irrigation and remote monitoring and controlling. Architecture of this system has basically three layers: sensing layer, processing layer, and application layer. Sensing layer deals with

Voyage of IoT in the Ocean of Technology  21

Figure 1.7  Future smart farm [28].

functionalities of different sensors such as soil moisture sensor and water level sensor that receive information from surrounding and pass to its cluster head. Such received information transfer from cluster head to remote server for further processing and analytics will be done at different application sides on such processed data to get the ultimate result. Such analytics results decide what should be the next step to follow and accordingly sends signal to/stop signal from actuators (e.g., water pump motor) to actuate (start/stop). ✓✓ In other agriculture domains, IoT applications such as environment monitoring systems will sense the environmental data such as level of carbon dioxide, level of nitrogen, and level of oxygen in the surroundings and alert if it goes beyond the lower level. At this time, it checks crop-based requirements accordingly and informs the remotely existed farmer community so that they can take action accordingly. ✓✓ In automated seed sowing systems, there are sensor mounted tractors that can monitor the shift of the tractor and accordingly dig soil and another sensor pushes seed into the soil. So, using such sensor-based sowing automation, the farmer community can get proper inline and depth seed sowing that can be easily maintained during its production phase and thereby increasing the overall production of the related crop. ✓✓ Soil fertility monitoring systems will basically consist of different sensors that can sense different micronutrients from soil. In its processing part, it compares with related crop ideal requirements, and if a gap is found beyond threshold, it sends an alert to a remotely existing famer on his smart device. So, by utilizing IoT applications in the agriculture domain, specifically at farm level as mentioned by above various IoT-based applications, our traditional farm can act as a Smart Farm that helps the farmer community to increase crop production quality and quantity and thereby achieve their overall goal of profit making with less sweat.

22  The Smart Cyber Ecosystem for Sustainable Development

Conclusion This chapter explains the Introduction of IoT and its basics. It covers Technological Evolution and Associate Technology such as IoT network and communication protocols. In the voyage of this chapter, it also explains the “Interoperability” as a solution of the serious issue of heterogeneity in IoT. This chapter moreover discusses some practical aspects in IoT programming using Arduino, Raspberry Pi, and Python programming language. It also explains IoT applications with a splendidly useful application of IoT in the agriculture domain such as “Smart Farm”. The main aim of this chapter is to draw attention and interest of the reader toward the IoT domain and induce him/her toward this domain which may result into innovative ideas in this domain.

References 1. Tripathy, B.K. and Anuradha, J. (Eds.), Internet of Things (Iot) Technologies, Applications, Challenges, and Solutions, pp. 41–59, CRC Press, Taylor and Francies Group Ch-3, London, 2018. 2. Patel, K.K. and Patel, S.M., Internet of Things-IOT: Definition, Characteristics, Architecture, Enabling Technologies, Application & Future Challenges. Int. J. Eng. Sci. Comput., 6, 5, 6122–6131, May 2016. 3. Vermesan, O., Friess, P., Guillemin, P., Gusmeroli, S. et al., Internet of Things Strategic Research Agenda, in: Ch.02 - Internet of Things-Global Technological and Societal Trends, River Publishers, Denmark, 2011. 4. Internet of Things: Evolution and technologies from a security perspective, vol. 54, p. 101728, Elsevier - Sustainable Cities and Society, Manchester, U.K., March 2020, https://doi.org/ 10.1016/j.scs.2019.101728. 5. Silva, B.N., Khan, M., Han, K., Internet of Things: A Comprehensive Review of Enabling Technologies, Architecture, and Challenges, pp. 205–220, Taylor and Fransis Online Published online, Howick Place, London, 08 Feb 2017, https://doi.org/10.1080/02564602.2016.1276416. 6. Learning Internet of Things, in: Chapter – 06 The XMPP Protocol, pp. 125–162, Peter Waher, PACKT Publishing, Birmingham – Mumbai, 2015. 7. Learning Internet of Things, in: Chapter – 05 The XMPP Protocol, pp. 107–123, Peter Waher, PACKT Publishing, Birmingham – Mumbai, 2015. 8. The Evolution of the Internet of Things, White Paper, Jim Chase, Texas Instruments, Dallas, Texas, September 2013. 9. Poole, I., IEEE 802.15.4 Technology and Standard, online: URL: https://www.radio-electronic. com/info/wireless/ieee-802-15-4/wireless-standard-technology.php. 2013. 10. Fenzel, L., Difference between IEEE 802.15.4 and Zigbee online URL: htts://www.elecronic­ design.com/what-s-different-between-ieee-802154-and-zigbee-wireless. 11. Agarwal, T., Zigbee wireless technology Architecture and Application. Online URL https:// https:elprocus.com/what-is-zigbee-wireless-technology-architecture-and-its-application. 12. Acosta, G., The Zigbee Protocol online URL: https://www.netguru.com/codestories/the-zigbeeprotocol. 13. Shelby, Z., Hartke, K., Bormann, C., The Constrained Application Protocol (COAP), in: Internet Engineering Task Force (IETF), Standard Track, Bremen, Germany, 2014. 14. Tezer, O.S., An Advanced messaging queuing protocol walkthrough, Digital Ocean–Online, 2013.

Voyage of IoT in the Ocean of Technology  23 15. Sulthana, M.R., A Novel Location Base Routine Protocol for 6LoWPAN, a developer cloud environment. 16. Feng, A., A Survey of Protocols and Standards for Internet of Things. Adv. Comput. Commun., 1, 1, pp. 1–21, 2011. 17. Egan, M., What is NFC? How to use NFC on your smartphones, Techadvisor, (online), Euston Road, London, 2015. 18. Tutorialspoint – Online, Wireless Communication – Bluetooth. 19. Čolakovića, A. and Hadžialićb, M., Internet of Things (IoT): A review of enabling technologies, challenges, and open research issues, vol. 144, pp. 17–39, Elsevier - Computer Networks, Amsterdam, The Netherlands, 24 October 2018. 20. Gomathi, R.M., Hari Satya Krishna, G., Brumancia, E., Mistica Dhas, Y., A Survey on IoT Technologies, Evolution and Architecture. 2nd International Conference on Computer, Communication, and Signal Processing (ICCCSP 2018). 21. Arduino Store, Online https://store.arduino.cc/usa/arduino-uno-rev3 (image download). 22. Arduino Integrated Development Environment-IDE https://www.arduino.cc/en/Main/ Software 23. Learning Internet of Things, in: Chapter – 01 Preparing our IoT Project, pp. 11–33, Peter Waher, PACKT Publishing, Birmingham – Mumbai, 2015. 24. Raspberry Official Website - https://www.raspberrypi.org/downloads/ (for image and software) 25. Official Web site – Python - https://www.python.org (for documentation and download) 26. Vermesan, O., Friess, P., Guillemin, P., Gusmeroli, S. et al., Internet of Things Strategic Research Agenda, in: Ch.03 - Internet of Things-Global Technological and Societal Trends, River Publishers, Denmark, 2011. 27. Le, D.-N., Van Le, C., Tromp, J.G., Nguyen, G.N. (Eds.), Emerging technologies for health and medicine_ virtual reality, augmented reality, artificial intelligence, internet of things, robotics, industry 4.0, Scrivener Publishing, Wiley. October 2018. 28. Agriculture 3.0 or (smart) agroecology, online: https://www.grain.org/en/article/6280agriculture-3-0/-or-smart-agroecolog 29. AgriSens:Development of Sensor based Networking System for Improved Water Management for Irrigated Crops, A Project Funded By MHRD, Govt. of India, Undertaken by SWAN, department of CSE & AGFE, IIT Kharagpur, official website: www.iitkgp.ac.in 30. http://www.gartner.com/it-glossary/internet-of-things/ - visited on 22/07/2020 at 5.30 p.m. 31. https://en.oxforddictionaries.com/definition/sensor - visited on 26/07/2020 at 4 p.m. 32. https://businessdictionary.com/definition/sensor.html - visited on 26/07/2020 at 4.30 p.m. 33. http://www.electronics-tutorials.ws/io/io_1.html - visited on 26/07/2020 at 4.30 p.m.

2 AI for Wireless Network Optimization: Challenges and Opportunities Murad Abusubaih

*

Department of Electrical Engineering, Palestine Polytechnic University, Hebron, Palestine

Abstract

Nowadays, Artificial Intelligence (AI) and Machine Learning (ML) are gaining increased attention. The huge amount of information coupled with a plethora of multimedia applications have posed a great challenge to scientists and engineers to handle the big data and manage various resources. All of this prompted researchers to think of innovative ways to make best use of AI and its tools to address existing and emerging problems in the field of data science and data networks. This had an impact on developing the concept of self-organized networks and systems. This chapter discusses a state-of-art of AI concepts and tools applied to wireless networking. We firstly introduce the AI concepts. We review self-organizing and cognitive networks. Then, we introduce the ML approach. We discuss how AI and ML would contribute to the management of wireless networks as well as the optimization of their operation. To help researchers gain a focused knowledge on the role of AI concepts in facilitating solutions to various problems in wireless networks, we discuss different areas and challenges where AI and ML have been used effectively to overcome those challenges. Keywords:  Artificial intelligence, machine learning, wireless networks, cognitive networks

2.1 Introduction to AI Artificial Intelligence (AI) is a field of science that is constantly evolving and accelerating. It has recently witnessed great momentum in being one of the scientific fields that have become affecting all sciences. AI has transformed the research path to new directions in order to provide effective solutions to many problems facing all science and engineering fields. In fact, the concepts of AI go back to the 1940s and 1950s, when scientists from different disciplines explored the possibilities of artificial brains and defined machine intelligence. AI systems can be divided into three types: analytical AI, human-inspired AI, and humanized AI. Analytical AI has characteristics compatible only with cognitive intelligence, where new knowledge and decisions are generated through learning and based on previous experience. Human-inspired AI is a mix of cognitive and emotional intelligence. Email: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (25–56) © 2021 Scrivener Publishing LLC

25

26  The Smart Cyber Ecosystem for Sustainable Development Human emotions are understood, in addition to the cognitive elements, and then used in determining decisions. Humanized AI uses cognitive, emotional, and social intelligence, capable of being self-conscious, and self-aware in interacting with others. The basic idea of AI is based on a simulation process of the interaction of data in human thinking, trying to understand human intelligence and then developing intelligent machines. AI has the ability to access objects, categories, their characteristics, and the relationships between them in order to apply knowledge engineering. AI aims to expand the capabilities of mankind in carrying out various tasks and consolidate the principles of intelligence in machines and devices in order to save time and effort and to provide distinguished services in various fields. Nowadays, we are witnessing the emerging of many smart devices in different fields, especially in engineering and medical sciences. Specific examples are computer vision, natural language processing, the science of cognition and reasoning, robotics, game theory, and machine learning (ML). Intelligent machines would have some of the capabilities related to human thinking in dealing with problems and make appropriate decisions for any event that may appear during machine operation. It is known that existing networks lack the intelligence needed to support future nextgeneration networks that are expected to be self-adaptive. Mobile networks consist of a large number of elements that interact with each other, creating a great complexity in the system that operates these elements together. Wireless networks constitute one of the most important areas that aspire to benefit and consolidate the principles of AI in order to adopt solutions to many problems appeared previously and appear currently in this field. Although we observe a great revolution in scientific research that relies on AI tools to develop and design wireless networks, applying AI approaches to network planning, design, and operations is still in the early stages. This is due to the fact that existing network architectures are not suited to the AI-enabled networks. Researchers are looking not only at the use of AI-based solutions to current problems, but noticeable research have returned to previous problems and tried to develop AI-based solutions. Later in this chapter, we will discuss recent research issues that can benefit from and exploit the principles of AI and ML. The main research directions that use the AI paradigm are as follows: • Expert Systems An expert system is a software system that relies on human expertise for decision-making. It is appropriate to deal with problems that involve incomplete information or big data. • Machine Learning ML relies primarily on how the computer simulates the behavior of human learning, then restructures the knowledge and acquires new skills to continuously improve performance. • Pattern Recognition The concept of pattern recognition is applied to process monitoring that assumes a relationship between data patterns. The research in pattern recognition includes two main issues: the first relates to object perception and the second relates to determining the category to which the object belongs. • Neural Networks The concept of artificial neural networks is based on non-linear mapping between the system’s inputs and outputs. It consists of interconnected neurons

AI for Wireless Network Optimization  27 arranged in layers. The layers are connected, allowing signals to propagate from the layers’ inputs across the network. A neural network stores data, learns from it, and improves its capabilities to sort new data. • Deep Learning Deep learning is the application of the concept of artificial neural networks to learning tasks that contain more than one hidden layer. It is part of a larger group of ML techniques that are based on representations of learning data. Deep learning concepts come from artificial neural network research, which opened a window to a new field of ML. Concepts of deep learning have been applied in various fields including computer vision, speech recognition systems, natural language processing systems, voice recognition systems, social networking systems, automatic translation systems, and bioinformatics systems, where the adoption of deep learning techniques has led to more effective results as compared to human experience and previous systems.

2.2 Self-Organizing Networks The primary goal of mobile networks is to connect mobile phone users together as well as to the Internet. Therefore, wireless network operators install large number of base stations or access points in the regions that will be covered. Each base station or access point covers a specific geographical area called a cell. Mobile networks allow users to transparently move between cells via a process called handover. Network users wish that the service provided to them is uninterrupted, whether with regard to the quality of phone calls or the speed with which they surf the Internet. Self-Organizing Networks (SONs) is an evolving technology used to automate planning, configuration, optimization, and healing of networks. SON is included as part of the mobile networks standards such as such as Long Term Evolution (LTE). The quick evolution in wireless network industry have led to parallel operation of 2G, 3G, 4G, 5G, and emerging 6G networks that need to be managed and controlled with minimal human effort. SON is a promising technology to realize solutions for the control and management of this heterogeneous network regime. The technology suggests a set of concepts to automate network management toward a goal of improving quality of service (QoS) and reduce burdens of networks management on network administrators [1].

2.2.1 Operation Principle of Self-Organizing Networks With SON, network administrators predefine a set of key performance indicators (KPIs) regarding QoS and other operational functions. Then, the network uses modules and algorithms to self-monitor and optimize its parameters, trying to achieve the predefined KPIs. This is considered as a closed loop control process, by which a network gains understanding of the operation environment and users’ behavior and adapts its parameters accordingly to achieve the intended performance goal, but at same time avoid any misconfiguration of parameters that may lead to service disturbances [2]. In the following subsection, we elaborate more on the features of SONs illustrated in Figure 2.1.

28  The Smart Cyber Ecosystem for Sustainable Development Configuration Parameters

Wireless Network

Measurements

SelfConfiguration

Measurements Self-Healing Configuration Parameters SelfOptimization

Measurements Configuration Parameters

Figure 2.1  SON features.

2.2.2 Self-Configuration Mobile communications networks are heterogeneous networks comprised of multiple technologies, such as LTE, EDGE, and UMTS. The number of mobile users is incredibly increasing which makes the installation and configuration of base stations a tedious process. Therefore, self-configuration is a process that reduces the time required for these tasks. Self-configuration provides an initial setup of the network elements. It consists of three stages. The first stage relates to automatic connection to the network, security procedure, and establishing a secure connection between network elements and the network core. The second stage is the programming of network elements, while the third stage relates to the configuration of radio parameters.

2.2.3 Self-Optimization Mobile networks are dynamic in nature. This pertains to traffic characteristics, the volume and variability of data exchanged between network elements, the joining of new users, the leave of others, and the movement of users among network cells. This results in variations of network performance as well as the level of service that users are experiencing. Therefore, self-optimization aims to maintain an optimal performance level for all network elements, through analysis of data measured and exchanged by network elements.

2.2.4 Self-Healing The larger the network size, the more likely that failures will occur. The objective of self-healing is to continuously monitor the network in order to automatically detect and recover from unexpected possible failures. In future networks, it is expected that self-healing enables the network to predict faults and automatically take the necessary measures to avoid service degradation and disruptions.

AI for Wireless Network Optimization  29

2.2.5 Key Performance Indicators KPIs are simple indicators that represent network performance. Here, we present examples of some important indicators: • Channel Quality Indicator: This represents the connection quality to all users in a cell. Obstacles and multipath fading are major factors that impact channel quality. • Handover Rate Indicator: This represents the mobility pattern of network users. It indicates the signaling traffic on the backbone network units which affects the overall network performance. • Cell Load Indicator: This represents the amount of load on a cell, in terms of users, traffic load, or a cost function. • Quality of Experience (QoE): This represents the satisfaction level of all users in the network or within each cell. Such indicator would characterize the QoS level users are experiencing.

2.2.6 SON Functions It is important to discuss the fundamental optimization tasks of SONs. In this section, we present some important tasks: • Coverage: Coverage optimization is a process through which a network tries to cover an intended area with minimal number of base stations and transmit power levels. • Capacity: Capacity optimization refers to the process of providing users with the best possible QoS using minimal radio resources. This would imply radio frequency assignment and interference mitigation techniques. • Mobility: Mobility optimization deals with the process of ensuring transparent user movement between cells and at the same time minimizing the number of unnecessary handover requests. • Load Balancing: This refers to the process of distributing the load among network base stations, trying to maximize the QoE in the network and minimize the overhead on core network elements.

2.3 Cognitive Networks Nowadays, communication networks are getting more complex and their configuration and management to achieve performance goals have become a challenging task. This is due to the following: • • • •

The significant increase in the number of network users. The increase of the number of required networking elements at the network core. The huge number of mobile applications. The diversity of traffic.

30  The Smart Cyber Ecosystem for Sustainable Development The idea of cognitive networks is to improve the performance of networks and reduce the effort required for their configuration and management. Unlike current technologies, in which networking elements are unable to make intelligent decisions, the elements of a cognitive network have the ability to learn and dynamically self-adjust as response to changing channel and network conditions. Cognitive network elements utilize the principles of logic and learning in order to improve performance. Decisions are made to improve the overall network performance, rather than the performance of individual network elements. Thus, cognitive networks achieve the goal of intelligent, self-adjustment, and improved network performance, by intelligently finding optimal values of many adjustable parameters. They are required to learn the relationships among network parameters of the entire protocol stack. As we indicated, a cognitive network should provide better performance to users. The cognition can be used to improve: utilization of network resources, QoS, security, access, control, or any other issue related to network management. It must be emphasized that cognition is not only related to wireless networks, but also the idea applies to the management of network infrastructure and the various network elements [3]. To stimulate transition to cognitive networks, their performance must outweigh all additional complexities that they require. The question is how to measure the cost of a cognitive network. Such cost would primarily depend on the communications required to apply cognition, the architecture complexity, maintenance cost, and the operational complexity. For example, in wired networks, user’s behavior is clear and easily predictable, and therefore, it may not be interesting for some people to employ cognition with this type of networks. On the contrary, wireless networks often include heterogeneous elements and have characteristics that cannot be easily predicted, making them the best candidates to adopt the cognition concept. Cognitive networks should use different measures, tools, and patterns as inputs to the decision-making processes. Then, they come up with results in the form of procedures or commands that can be implemented in modifiable network elements. It is important to note that the cognitive network must adapt to changes in the environment in which it operates and anticipate problems before they occur. Their architecture must be flexible, scalable and be supportive of future improvements and extensions. Several research studies have been discussing the architecture and functionalities of cognitive networks. There is a need to rethink about network intelligence from being dependent on resource management to understanding the needs of network users and then transferring intelligence also to the elements of the network. The central mechanism of the cognitive network is the cognitive process. This process implements real learning and decides the appropriate responses and actions based on observations in the network. The operation of the cognitive process mainly depends on whether its implementation is central or distributive as well as on the amount of state network information.

2.4 Introduction to Machine Learning ML is a subset of AI. The aim of ML is to develop algorithms that can learn from data and solve specific problems in some context as human do [4]. ML has been proving its ability

AI for Wireless Network Optimization  31 to overcome the challenges and complexities of mathematical formulation and solution of complex problems, including wired and wireless networking problems that require effective methods to quickly respond to dynamical changes of channels as well as the increasing diversification of services. Dynamic ML algorithms are able to process data and learn from it. They are replacement of complex algorithms which are written in a fixed way to conduct specific tasks. The basic concept of ML is through training data that is used as input to the learning algorithm. The learning algorithm then produces a new set of rules, based on inferences from data, which results in a new algorithm. The new algorithm is officially referred to as the ML model. Traditional algorithms are comprised of a set of pre-programmed instructions used by the processor in the operation and management of a system. However, instructions of ML algorithms are formed based on real-life data acquired from the system environment. Thus, a machine is fed a large amount of data, it will analyze and classify data, then use the gained experience to improve its own algorithm and process data in a better way in the future. The strength of ML algorithms lies in their ability to infer new instructions or policies from data. The more data is available for the learning algorithms during the training phase, the more ML algorithms will be able to carry out their tasks efficiently and with greater accuracy.

2.4.1 ML Types Depending on the type of tasks, there are two types of ML: • Regression Learning It is also called prediction model, used when the output is a numerical value that cannot be enumerated. The algorithm is requested to predict continuous results. Error metrics are used to measure the quality of the model. Example metrics are Mean Absolute Error, Mean Squared Error, and Root Mean Squared Error. • Classification Learning The algorithm is asked to classify samples. It is of two subtypes: binary classification models and multiple classification models. Accuracy is used to measure the quality of a model. The main difference between the algorithms for classification and regression is the type of output variable. Methods with quantitative outcomes are called regressions or continuous variable predictions. Methods with qualitative outputs are called classifications or discrete variable predictions.

2.4.2 Components of ML Algorithms A formal definition of a ML algorithm is “A Computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks T, as measured by P, improves with experience E” [5]. • Tasks: A task defines a way to process an object or data. An example task is classification, which is a process of assigning a class label to an input object

32  The Smart Cyber Ecosystem for Sustainable Development or data point. Regression is another task example, which involves assigning a real value to an object or data point. • Performance Measure: Defines the criteria by which a ML algorithm is evaluated. In classification algorithms, accuracy refers to the percentage of correct assignment of class labels to objects or data points. Normally, data is divided into two sets. The first is used for training, while the second is used for testing. • The Experience: It refers to the knowledge that a ML gains while learning. It divides the ML algorithms into the types explained in the next subsection.

2.4.3 How do Machines Learn? Intelligent machines learn from the data available in their environment. The process of applying ML consists of two phases: The training phase and the decision-making phase. In the training phase, ML techniques are used to learn the system model using training dataset. In the decision-making phase, the machine shall be able to estimate the output for each input data point using the trained model. According to the training method, ML techniques can be classified into four general types. Many advanced ML techniques are based on those general types. Figure 2.2 illustrates these types.

2.4.3.1 Supervised Learning This learning method requires a supervisor that tells the system what is the expected output for each input. Then, the machine learns from this knowledge. Specifically, the learning algorithm is given labeled data and the corresponding output. The machine learns a function that maps a given input to an appropriate output. For example, if we provide the Supervisor/Targets

Error Input

Supervised Learning

Output

Input

Unsupervised Learning

Output

Process Reward Input

Figure 2.2  Machine learning types.

Performance

Reinforcement Learning

Output

AI for Wireless Network Optimization  33 H1

X2

H2 (Optimal Hyperplane) H3

Margin Support Vectors X1

Figure 2.3  Illustration of SVM.

ML system during the training phase with different pictures of cars, and with information indicating that these are pictures of cars, it will be able to build a model that can distinguish the cars’ pictures from any other pictures. The quality of a supervised model depends on the difference between the predicted output and the exact output. The convergence speed of supervised learning is high although it requires large amount of labeled data [6]. Next, we discuss the well-known supervised learning algorithms. Support Vector Machine Support Vector Machine (SVM) algorithm is a linear supervised binary classifier. It separates data points using a hyperplane. The best hyperplane is the one which results in maximum separation between the two given classes. It is called the maximum margin hyperplane. SVM is considered to be a stable algorithm applied for binary classification. For multiple classification problems, the classification tasks must be reduced to multiple binary classification problems. The basic principle of SVM is illustrated in Figure 2.3. K-Nearest Neighbors A non-parametric learning algorithm used for classification and regression. The algorithm does not require any assumption on the data distribution. The objective of KNN is to decide the class of a data point based on the results of majority voting of its K-nearest neighbors (KNNs). The idea is that a data point is identified to belong to a certain class if KNNs belong to that class. A weight can be used for each neighbor that is proportional to the inverse of its distance to the data point in the classification process. KNN is easy to implement, not sensitive to outliers, highly accurate, and easily calculates features. It is also suitable for multi-class classification applications. The basic principle of KNN is illustrated in Figure 2.4.

2.4.3.2 Unsupervised Learning In this technique, data is submitted to the learning algorithm without predefined knowledge or labels. Thus, the machine has to learn the properties of the dataset by itself through the study of unlabeled training data. The algorithm shall be able to define patterns from the input data. Observations are clustered in groups according to the similarities between them. The clustering algorithm examines the similarity of observations based on their features.

34  The Smart Cyber Ecosystem for Sustainable Development Class 1 Data

Test Sample

K=3

Class 2 Data K=5

Figure 2.4  Illustration of KNN.

Observations are then grouped in a way that puts elements that share a high similarity in the same group. Normally, algorithms use distance functions to measure similarities of observations. With Unsupervised learning, no prior knowledge is required. However, this comes at the cost of reduced accuracy [6]. The most commonly known unsupervised algorithm is clustering. Clustering algorithms divide data samples into several categories, called clusters. Clustering algorithms are of four main types [7]: • Centroid-Based Clustering: Clusters are defined using centroids. Centroids are data points that represent the proto-element of each group. The number of clusters has to be defined beforehand and is fixed. In the beginning, cluster centroids are defined randomly and will be shifted in the state space iteratively until the specified distance function is minimized. • K-Means Clustering is a simple and most common centroid-based method. The objective is to partition data points into K clusters, where each data point should belong to the cluster with the nearest mean. Initially, K mean points are randomly picked (the centroids). Then, the algorithm iterates on each data point and computes the distance to the centroids. The data point is judged to belong to the point to which the computed Euclidean distance is minimum. Thus, the method minimizes the distance between points and their corresponding centroids. Centroids are updated based on their assigned data points. The process continues until the centroids do not change. Figure 2.5 illustrates the concept of K-means clustering. • Hierarchical Clustering: In this type, the number of clusters is not defined a priori; rather, it is iteratively increases or decreases. In the beginning, all observations are included in one cluster. Then, the cluster is split according to the largest distance between the data points. Once a sufficient number of clusters is reached, the process is stopped.

AI for Wireless Network Optimization  35 X2

Initial Means Final Means

X1

Figure 2.5  Illustration of K-means clustering.

• Density-Based Clustering: In this type of clustering, the algorithm tries to find the areas with high and low density of observations. Data points that are within a specified distance become centers of a cluster. Other data points either belong to a cluster border or considered as noise.

2.4.3.3 Semi-Supervised Learning This learning approach combines both supervised and unsupervised ML techniques. Thus, the machine learns from both labeled and unlabeled data. This approach is more realistic for many applications, wherein small amount of labeled data is available, but the collection of large set of labeled data is not easy or impractical.

2.4.3.4 Reinforcement Learning Similar to unsupervised learning in the sense that the machine has to learn by itself. However, a reward mechanism is applied to tune the algorithm based on observation of performance, enabling continuous self-update of the machine. Reinforcement learning algorithms try to define a model of the environment by determining the dynamics of the environment. The algorithm uses an agent which interacts with a dynamic environment in a trial-and-error manner. It provides feedback to the algorithm. The agent makes decisions on what actions to be performed to optimize the reward. A policy determines how the agent should behave at a given time. Thus, the algorithm learns by exploring the environment and exploiting the knowledge. The feedback from the environment is used to learn the best policy to optimize the cumulative reward. The most commonly known reinforcement algorithm is the Q-Learning. The RL algorithm interacts with the environment to learn Q values. The Q value is initialized. The machine observes the current state, chooses an action from a set of possible actions, and performs the action. The algorithm observes the reward and the new state. The Q-value is updated based on the new state and the reward. Then, the state is set to the new state and the process repeats until a terminal state is reached.

36  The Smart Cyber Ecosystem for Sustainable Development

2.4.4 ML and Wireless Networks It is expected that future wireless networks will be highly integrated and a qualitative change will occur regarding the use of high frequencies and wide channels. In addition, the networks are expected to run a large number of base stations and serve high density of users. Future communication networks are dynamic and may also be without cells and massive-MIMO. They will be intelligent, flexible, and highly resilient [8]. ML is a promising tool for efficient management of future wireless networks.

2.5 Software-Defined Networks Current networks are characterized by their distributed nature, as each node (router/ switch) has the ability to view and act on the system partially and locally. Thus, global learning from network nodes that have a holistic view on the system will be very complicated. Further, current network designs impose significant limitations on network performance, especially under high traffic conditions. Consequently, the increasing demand for reliable, fast, scalable, and secure networks can adversely affect the performance of existing network devices due to the need to deliver a large volume of data both in the network infrastructure and devices. Current network devices lack the flexibility to handle different types of packets that may carry different contents due to the basic implementation of hardwired routing rules. In addition, the networks that form the backbone of the Internet must be able to adapt to the changing conditions without needing much effort for hardware and software adjustments. In order to reach a solution to the above discussed limitation issues, the rules for data processing must be implemented through software modules and not embedded in the hardware. This approach enables network administrators to have more control over the network traffic, and thus can greatly improve the network performance and effectively use the network resources. This innovative approach is called Software-Defined Network (SDN) [9]. SDN was released as open source software in 2008 with the OpenFlow project at Stanford University. It decouples the control and data planes in routers and switches, allowing the underlying infrastructure to be separated from application and network services. Thus, SDN separates the decision-making modules about where traffic is sent [the control plane (CP)] from the underlying systems responsible for forwarding the actual traffic (the data plane). Network resources are managed by a centralized controller which performs as the network operating system. The controller can dynamically program the network in real time. It collects information about network status and operation details. Therefore, the controller can globally detect available network resources and requirements. This paradigm creates a global view of the entire network, enabling global automatic management and control without needing to configure devices individually. The SND technology has several advantages: • Efficient utilization of network resources. • Enables development of programming-based solutions for network configuration and management. • Provides a perfect ecosystem for ML paradigm and intelligent applications.

AI for Wireless Network Optimization  37 • Simple and improved network management, control, and data manipulation, since network administrators can remotely alter the network configuration and operation as response to dynamical changes in the network. • High speed, through optimal handling of the traffic load. • Adopts the virtualization technology, which allows running multiple applications over the same shared hardware. Combining AI and SDN has been attracting researchers in recent years to develop network management and operation mechanisms. The SND architecture provides centralized control of network policies and enables administrators to effectively overcome problems with ML methods.

2.5.1 SDN Architecture The architecture of SDN is comprised of three planes: • Data Plane: comprised of the forwarding devices, i.e., switches and virtual switches. Unlike distributed network architectures, in which switching and routing devices listen to events from neighboring network elements and make decisions based on a local view, switches and routers are responsible for forwarding, dropping and modifying packets based on policies received from the CP. • Control Plane: The CP is considered to be the brain of SDN. It can program network resources and dynamically update the rules of forwarding, in addition to making the management of the network flexible through the centralized controller. The centralized controller controls communication between switches and applications. On the other hand, the controller exposes the network status and summarizes the information to the application plane. Also, the CP translates the requirements from applications to specific policies and distributes them to devices. Further, it provides the basic functions needed by most network applications such as routing algorithms, network topology, device configuration, and state information notifications. • Application Plane: composed of network applications that define management and optimization policies to be applied on the network. Applications can get network state information from the controller and implement the needed control to change network behavior. The inclusion of ML in SDN may require a new architectural structure that differs from the traditional of SDN. In [10], a new plane is proposed called the knowledge plane KP as shown in Figure 2.6. The KP hosts ML algorithms that use statistical learning to learn the network behavior. These algorithms contribute to decision-making. Hence, the KP in SDN communicates directly with the controller, which, in turn, asks the network elements to implement decisions. The controller gets information from network devices through the OpenFlow protocol. A server is used to process information and run ML algorithms. The execution of recommended commands is the responsibility of the controller which is connected to the KP. On the top, the application plane is running to manage the network.

38  The Smart Cyber Ecosystem for Sustainable Development Applications Application Plane

Knowledge Plane

Control Plane

Data Plane

Figure 2.6  The SDN architecture with knowledge plane.

2.5.2 The OpenFlow Protocol The open source OpenFlow protocol enables the realization of the SDN technology. It has been used for researching different protocol designs over existing hardware. OpenFlow is widely used for the communication between the control and the data planes. It is developed by the Open Networking Foundation (ONF). It is the interface between devices and the SDN controller, providing the rules for switching control features from network devices to the central controllable software. It has a controller and a switch, functioning as secure channel as shown in Figure 2.7. The controller can modify, discard, and send packets to the switch. The path of the packet is determined at the times of packet transmission. OpenFlow calculates the path and sends it to the switch, which stores it in the Flow Table. When a switch receives a packet, it looks up the flow table and sends it along the stored path [11]. The primary task of the switch is to exchange data using flow tables, which are controlled by

OpenFlow Switch OpenFlow Software Hardware

Network Devices

Figure 2.7  The OpenFlow architecture.

Secure Channel

Flow, Group Tables

Controller

AI for Wireless Network Optimization  39 the controller of the CP. This architecture simplifies the design of switches and reduces their tasks, because they have become just data deliverers, without being required to perform any of the control functions. The implementation of an SDN controller can be centralized or distributed. In the centralized implementation, a single SDN controller centrally controls and manages all network devices, which would possibly lead to bottleneck. Distributed implementation of the SDN controller would overcome this issue. The CP may include multiple controllers, depending on the network size. This will help boosting the network performance.

2.5.3 SDN and ML SDN has strengthened applying programmatic principles on network, allowing network administrators to have precise, flexible, and innovative control of the network and thus reducing operational expenses. The SDN architecture provides an opportunity to more efficient application of cognitive network concepts in a centralized system, leading to self-aware networks. The adoption of SDN-based systems highly depends on their success in providing solutions to problems that could not be solved by traditional network architectures and protocols [12]. Applying ML techniques with SDN is considered to be effective for the following reasons [13]: • The recent advanced developments in computing and the accompanying advanced processors, thus creating a new opportunity to apply promising learning techniques. • It is well known that ML algorithms depend on data. The SDN controller has a holistic view on the network and is able to collect different network data, simplifying the application of ML algorithms. • Based on the ability of the SDN to act in real time and deal with historical data, ML techniques can create intelligence in the controller unit, by conducting data analysis relying on analyzed data in decision-making and thus improving the network and its services. • The programmatically feature of SDN can help to find optimal solutions to network problems such as configuration and resource allocation. Thus, ML algorithms can be implemented in real time.

2.6 Cognitive Radio Networks The Federal Communications Commission (FCC) defines cognitive radio as: “a radio that can change its transmitter parameters based on interaction with the environment in which it operates”. The main features of cognitive radio are as follows [14]: • Awareness: CR is aware of its surrounding environment through the sensing capability.

40  The Smart Cyber Ecosystem for Sustainable Development • Intelligence: CR is a programmable intelligent wireless communication system capable of learning from information collected from the environment. • Adaptivity: CR adapts to the variation of the radio spectrum conditions and application requirements by dynamically reconfiguring its operational parameters. Dynamic Spectrum Access has been proposed for efficient utilization of radio spectrum. Spectrum bands are categorized as licensed and unlicensed bands. Licensed bands are used by licensed users, called Primary Users (PUs). They have the priority to use the spectrum. Unlicensed users, called Secondary Users (SUs) can use the licensed bands as long as the PUs are not temporally using it; or as long as the PUs’ can properly be protected. However, SUs should vacate the licensed bands immediately when a PU is detected to be active. This will significantly improve spectrum utilization. SUs detect the conceivable vacant bands, determine operational channel, and eventually adjust their parameters. Thus, efficient spectrum sensing techniques are key to the successful operation of CR networks. In CR systems, SUs should be able to [14]: • Sense the spectrum bands and determine the possible channels as well as activity of PUs. • Decide on the quality of available channels that satisfy users’ requirements. • Share the available channels with other SUs. • Avoid harmful interference to PU who is starting to use the any channel by vacating the channels PU just start operating on. The detection of PU’s presence is a major challenge in CR. This process needs complex sensing technologies. This complexity stems from the nature of the electromagnetic signals, the multipath fading, and the changing interference. Spectrum sensing for CR systems is a very interesting research area. Researchers try to develop quick and accurate methods to detect the PU’s activity. The transmission cycle of a SU can be represented as seen in Figure 2.8. The sensing process is performed periodically using one of the sensing methods. During the sensing time, the SU does not transmit data. After sensing, SU can decide to transmit on the same channel or switch to another channel, depending on PU presence. The transmission continues until the next sensing period. Clearly, the sensing time should be as short as possible, but at the same time enough for accurate sensing. Therefore, there is a trade-off between protecting the PU’s QoS and improving the QoS of SUs. t (i)

Sensing

T(i)

Data Transmission

Operation Time

Figure 2.8  The transmission cycle of SU.

t(i+1)

Sensing

T(i+1)

Data Transmission

AI for Wireless Network Optimization  41

2.6.1 Sensing Methods The fundamental objective of sensing methods is to increase the positive detection probability and decrease the false detection probability of PUs. This leads to more protection of PUs and at the same time more utilization of the available spectrum. In this section, we present the main sensing methods. However, several other methods can be found in the literature. 1. Energy Detection The most commonly adopted method because of its simplicity and the associated computation overhead. It requires short sensing interval. An energy detector is used to detect a narrowband channel. The detected energy is compared to a predefined threshold. If the measured energy is found to be larger than the threshold, then PU is judged to be active. The selection of the threshold is a challenge because the noise level is normally unknown. 2. Cyclostationary Feature Detection This method aims at distinguishing between PUs’ signals, interference, and noise. This is achieved by identifying the cyclostationary features of signals, including modulation type, carrier frequency, and data rate. The implementation of this method needs sufficient prior information about these features of the PUs’ signals so that the method can use this knowledge as a base during the matching of measured features with those belong to PUs’ signals. Hence, sufficient number of samples is needed for accurate performance, leading to long sensing intervals. 3. Matched Filtering This method is considered to be the most accurate method that achieves higher detection probability in short sensing intervals. The basic idea of this method is that the sensed signal is passed thought a filter that is matched to the PUs’ signals. Despite its accuracy, the method is considered to be impractical in cases wherein PUs transmit signals of different features.

2.7 ML for Wireless Networks: Challenges and Solution Approaches In previous sections, we discussed various topics about recent trends in network configuration, control, and management. We indicated recent trends in network architecture design in order to respond to the requirements of large-scale networks. Large-scale networks feature a significant increase in the number of users and smart technology-based systems that have become an integral part of human life, especially with the vast amount of software applications that require high speeds and fast real time response. In addition to that, different applications emerged in the last years to allow easy and high quality communication between people. Such applications have become an important tool used even by the largest media stations. In the following sections, we review and discuss important challenges in wireless networking that can be better tackled, exploiting ML approaches. The application of ML in wireless networking aims at reducing human interaction and creating a self-driven network,

42  The Smart Cyber Ecosystem for Sustainable Development that are able to optimize and configure themselves. We focus on recent published research and try to shed light on important research aspects of the present and future. Our goal is to assist readers to identify the scientific areas and specific issues that need further research and exploration. We divide the discussion into three parts. In the first part, we focus on Cellular networks, while the second part focuses on wireless local area networks (WLANs). The third part is devoted to cognitive radio networks.

2.7.1 Cellular Networks 2.7.1.1 Energy Saving With the steady increase in the number of users of wireless networks and the need to deploy large number of base stations, and since base stations consume large energy; operating the network with minimum energy is a challenge. One way to reduce energy consumption is the idea of turning off some base stations if users can be served from others, while maintaining a reasonable QoS level. Learning the operation of the network over time helps in improving decisions about which base stations might be switched off. An SDN-based ML system for energy saving is proposed in [15]. Performance of neural networks and SVM algorithms is compared. The network trains itself using data collected from base stations and recommends the operator time periods during which some base stations are predicted to handle very low traffic and therefore can be switched off. The authors of [16] propose a Q learning method for base station on-off switching. The switching of base stations is defined as the actions, while the traffic load is defined as the state. The overall objective is to minimize energy consumption. Policy values are used by the controller to decide on switching. After performing a switch operation, the system state is changed and the energy cost of the former state is computed. If the energy cost of the newly executed action is smaller than energy costs with other actions, then the controller updates the policy value in order to increase the probability of selecting this action. With time, the optimal switching mechanism is obtained.

2.7.1.2 Channel Access and Assignment The effective use of wireless channels has become an urgent necessity, as many heterogeneous systems operate in the same frequency band. Thus, coexistence and organized access of the shared frequency chunks by systems are necessary. Consequently, any design of the wireless channel sharing mechanism should be based on a prediction of the behavior of networks users. In [17], the authors propose deep reinforcement ML-based MAC protocol for the coexistence of multiple heterogeneous networks. The method allows time-sharing access of the spectrum, by a series of observations and actions. The MAC protocol does not have to know the MAC mechanism of other networks and tries to maximize the throughput of all coexisting networks. The authors of [18] employ reinforcement learning for managing cell outage and compensation. The system state is constituted by the allocation of users to the resources of cells and the channel. Actions are related to the power control, while the rewards are quantified in terms of SINR improvement. The authors show that such ML-based approach provides improved performance.

AI for Wireless Network Optimization  43 In [19] and [20], the authors use clustering algorithms to group users that share common interests to reduce interference and collisions. The authors show that such clustering improves the access opportunities for wireless users. A cluster header (CH) is selected to collect data from all devices. It sends the data to base stations which schedule the transmission. Channel assignment is a well-known challenging issue in wireless networks, especially with systems of limited channels. Such systems highly suffer from interference, and the optimal selection of channels becomes important. Channel assignment problems are normally formulated as convex optimization problems, where algorithms needed to solve such problems are computationally complex. In a dynamic environment of wireless networks, understanding the behavior of network users and learning from previous data is expected to be a good approach for improving channel assignment mechanisms. The paper of [21] uses ML approach to tackle the channel assignment problem and developed a computationally efficient solution for this problem. The objective is to maximize the total data rate experienced by all users assuming limited resources and large number of network users. The convex optimization problem is converted to a regression problem. Ensemble learning is utilized to combine different machine learning models and improve the prediction performance.

2.7.1.3 User Association and Load Balancing User association and load balancing is a challenge that has been attracting researchers of wireless networks. The question is how to optimally assign users to base stations and distribute the load in a balanced way among network base stations. The aim is to achieve high QoS to all users and at the same time efficiently utilize network resources. The authors of [22] investigated the use of deep learning to perform user-cell association to maximize the total data rate in massive multiple input multiple output (MIMO) networks. The authors show how a deep neural network; that gets the geographical positions of users as input; can be trained to approach optimal association rule with low computational complexity. Association rule is updated in real-time considering mobility pattern of network users. A method for cell outage detection was proposed in [23] using neural networks and unsupervised learning. The main feature of the method is the training of the network which can be performed in advance even when the cell outage data is not available. Moreover, the developed method could work in time-varying wireless environments. The machine learns from measurement reports of signal power which are collected by mobile devices. The research work in [24] proposes a distributed, user-centric ML-based association scheme. The algorithm is based on fuzzy Q-learning, where each cell tries to maximize its throughput under infrastructure capacity and QoE constraints. With this scheme, cells broadcast data values to guide users to associate with best cells. The values reflect the possibility of a cell to satisfy a throughput performance level. Each cell tries to learn the optimal values through iterative interaction with the environment. In [25], the authors used realistic mobile network data and investigated methods for failure prediction. They compared the performance of the SVM and several neural networks.

44  The Smart Cyber Ecosystem for Sustainable Development

2.7.1.4 Traffic Engineering The process of analyzing traffic in networks is normally performed through examining messages and extracting information from them. This helps in developing effective assessment strategy of how network users behave and identifying their goals from using networks, as well as knowing the data paths and communication patterns. All of this can be used to provide information for network management algorithms and to optimize the use of network resources. Traffic engineering is related to two processes: Prediction and Classification. Traffic prediction is a process for anticipating the traffic volume based on previously observed traffic volume; while traffic classification is a process of identifying the type of traffic. The process of traffic classification is based on collecting large number of traffic flows and analyzing those using ML techniques. Classifying traffic would help in improving security, QoS, capacity planning, and service differentiation. Classes could be: HTTP, FTP, WWW, DNS, P2P, Skype, and YouTube. Classification can be based on one or more traffic parameters, such as port number, packet payload, host behavior, or flow features [26]. ML is considered as an efficient tool in [27] for applying traffic engineering concepts. The authors use naïve Bayes classification, which uses supervised learning to construct a learning model for traffic analysis and classification. They developed a new weight-based kernel bandwidth selection algorithm to improve the constructed kernel probability density and ML model. The authors of [28] developed and SDN-based intelligent streaming architecture which exploits the power of time series forecasting for identifying users’ data rate levels in wireless networks, trying to improve the QoS of delivering video traffic. The SDN architecture is comprised of Data Plane (Switching devices), QoE management plane (management, bandwidth estimator, monitor, policy enforcer, and bandwidth forecaster), and CP aims to support the delivery of video services and to provide the QoE-based resource allocation per user. The paper of [29] compares the performance of several supervised and unsupervised ML algorithms to classify traffic as normal or abnormal. In [30], the authors propose a traffic classification algorithm based on flow analysis. The algorithm is designed for SDN platforms. The work in [31] uses traffic classification as part of a traffic scheduling solution for a data center network managed by SDN. ML techniques are used to classify elephant traffic flows, which require high bandwidth. Then, the SDN controller uses classification results and implements optimization of traffic scheduling. The authors of [32] use two phases for detection of elephant flows using ML techniques in SDN-based networks. In the first phase, packet headers are used to distinguish between elephant flows from mice flows, low bandwidth flows. A decision tree ML algorithm is then used to detect and classify traffic flows. Also, the authors of [33] developed an OpenFlow-based SDN system for enterprise networks. Several classification algorithms were compared. An application of ML for improving the quality and latency of real time video streaming is proposed in [34]. The video quality is achieved through rate control, employing a DL-based adaptive rate control scheme. Two RL models are used. The first one is for prediction of video quality model, while the second is video quality RL. The predictor uses previous video frames to predict quality of future frames. The RL algorithm adopts and

AI for Wireless Network Optimization  45 trains the neural network based on historic network status and video quality predictions to decide rate control actions. In their research published in [35], the authors developed a method for traffic prediction based on the SDN architecture, where the controller gathers data and uses it to classify data flows into categories. Neural network algorithm is used to predict the expected traffic, leading to a system that can act to avoid traffic imbalance before it occurs.

2.7.1.5 QoS/QoE Prediction QoS parameters are normally used by network administrators to assess the network performance. The parameters include throughput, loss rate, delay, and jitter. However, QoE is a parameter used to represent the user perception and satisfaction of the services. Developing prediction methods for QoS and QoE parameters helps network operators and service providers to offer high quality services [13]. SDN has been used to facilitate the implementation of different algorithms for QoS/QoE prediction [36–39]. The authors of [36] propose a linear regression ML algorithm for QoS prediction in SDN-based networks. A decision tree approach is used to detect relations between KPIs and QoS parameters. The authors show that the method can predict congestion and thus provide recommendations on QoS improvement. The researchers in [37] utilize two ML techniques for estimating QoS parameters for video on demand applications. QoE prediction was addressed in [38–39]. The method of [38] was designed for video streaming in an SDN-based network, where QoS parameters are employed to estimate the mean opinion score. The SDN controller is used to adjust video parameters to improve QoE. In [39], the authors use neural network and KNN algorithms for predicting QoE parameters using video quality parameters.

2.7.1.6 Security Users only use secure networks. One major issue in networking is the attacks by intrusions. Detecting intrusion and responding to attacks is a real challenge, especially in wireless networks where data is communicated over a shared media. With the advent of ML technology, researchers have been trying to exploit ML techniques to overcome this problem. ML methods can process and classify traffic flows based on observable properties such as number of packets in a flow, flow duration, packet size, inter-packet arrival time, and flow size in bytes. Based on these properties, more advanced features can be computed. The authors of [40] propose a system for ML-based flow classification integrated in SDN. It exploits methods of extracting knowledge that can be used by the controller in order to classify flows. A supervised ML algorithm has been used for identifying the underlying application flow, while unsupervised learning algorithm has been used for clustering flows in order to identify unknown applications. The system is also able to detect groups of related flows and proved to detect anomaly and botnet, as well as honeypot traffic rerouting. The authors of [41, 42] show that employing user centric approaches combined with ML can improve the performance of anomaly detection in cellular networks. User centric approaches focus on the end user while developing designs and strategies for networks, thus the need of end users will tailor networking solutions. The study uses the SVM, KNN,

46  The Smart Cyber Ecosystem for Sustainable Development and an optimized version of decision tree, wherein algorithms learn and predict QoE scores for users. A node is judged to be dysfunctional if the maximum number of users connected to this network node have poor QoE scores. In [11], the authors developed an SDN-based system for real time intrusion detection using a deep learning-based approach. Data sets are used to train the ML algorithm, following the supervised learning approach. Then, a flow inspection module examines the flows and decides whether it is an intrusion flow or not. The SDN paradigm facilitates the implementation of the proposed method, as it provides means for designing flow-based monitoring and control mechanisms. A detailed intelligent system for an automated control of large-scale networks is developed in [43]. The system architecture exploits SDN and deep RL methods for intelligent network control. Among other objectives, the system can serve applications that require traffic analysis and classification. RL involves processes that learn to make better decisions from experiences by interacting with all network elements. The SDN architecture is comprised of three planes: forwarding plane, the CP, and the AI plane. The function of the forwarding plane is forwarding, processing, and monitoring of data packets. The CP connects the AI plane and the forwarding plane. The SDN controller manages the network through standard southbound protocols and interacts with the AI plane through the northbound interface. The AI plane generates policies. It learns the policy through interaction with the network environment. An AI agent processes the network state data collected by the forwarding plane, then transfers the data to a policy through RL that is used to make decisions and optimization. The researchers in [44] use KNN classification algorithm for detecting several types of attacks. The authors pointed out that with large training dataset, the computation of distances between the test point and training data is time-consuming as the algorithm needs also to sort and find the closest K neighbors. Author in [45] uses unsupervised ML for detecting anomalies in real networks. The proposed approach enables anticipation of anomalies before they become a real problem. The paper of [46] provides a detailed review of recent studies that combines ML and SDN technology to solve the intrusion detection problem. The authors compare the performance of supervised, unsupervised, semi-supervised, and DL algorithms.

2.7.2 Wireless Local Area Networks In recent years, we see tremendous widespread of WLANs, as they evolve to meet user’s requirements, especially the high speed Internet connection. Accurate prediction of WLANs performance is important for managing network resources. However, due to interference and the interactions between the physical and data link layers as well as the heterogeneity of WLAN devices, predicting and estimating the performance of WLANs is a difficult task. Many of the solutions use the Signal-to-Noise and Interference Ratio (SNIR) parameter. However, it has been proven that relying on this parameter to estimate the performance does not lead to satisfactory results. In fact, the performance of WLANs is more complex to be measured using SNIR, and it is a function of large number of interacting and related parameters that may change over time. A plethora of research studies has developed various solutions to different challenges based on the traditional architecture of WLANS, aiming to optimally exploit network

AI for Wireless Network Optimization  47 resources and provide high QoS to users. With the spread of AI systems and the tendency to develop AI-based solutions in all areas, researchers have been trying to adapt the ideas and tools of AI and study the possibility of using these tools for optimal operation and management of WLANs. In this section, we discuss some important studies in this field. ML and the combination of ML and SDN have been shifting the research in WLANs to a new direction what allows more practical solutions to complex networking problems. Such solutions do not only simplify the management of network but also alleviate the complexity of algorithms and facilitate reaching optimal operation of a dynamically changing network. The authors of [47] develop a framework solution for the control and management of WLANs based on the SDN approach. The system comprises a set of modules and ML algorithms, stressing the fact that modern and future WLANs will be intelligently controlled. The authors tried to show the strength of the developed solution by addressing the following issues: mobility, security, QoS, channel bandwidth allocation, coordinated transmission power, load balancing, and virtualization of wireless networks.

2.7.2.1 Access Point Selection Administrators of WLANs normally deploy large number of APs to cover an area and provide users best QoS. In such scenario, a user could be under the coverage of multiple APs and thus has the potential to select the AP to which it will connect. The selection of AP is important and affects the performance a user might experience in the network as well as the overall network performance. This is because wireless networks are highly dynamic, whereas the activation of a link between a user and an AP may influence other ongoing connections in same and/or neighboring cells. In current implementations of WLANs technology, a user selects an AP from which it gets the strongest signal during a scanning phase. It has been shown in large number of research studies that the legacy selection policy does not ensure best QoS for network users. Obviously, an AP to which a user has the strongest connection might be serving large number of users; hence, its cell will be highly congested [48]. The authors of [49] propose an SDN-based AP selection scheme for WLANs. The selection is based on the analysis of achievable throughput a user might get from potential cells. The system computes the throughputs that capture the channel competition among neighboring cells. Some cells may obtain few chances to get the channel. Even in the same cell, mobile users compete with each other for channel access. The authors noticed that the airtime completion and airtime share among WLAN users play a fundamental role in determining the QoS the user will get. The authors implemented their proposed method in an SDN framework comprised of three planes: data plane, CP, and service plane. The data plane consists of a number of thin APs that are responsible for data forwarding. The CP contains a SDN-WiFi controller, while the service plane contains a set of applications such as association control, load balance, and seamless mobility management. The proposed AP selection scheme is implemented in the service plane. The applications are installed on the SDN controller, which collects necessary information from networking devices and decides the best cell for each of newly joining users. In [50], the authors developed ML-based methods for detecting causes of unnecessary active scanning in WLANs. The authors argue that ML provides the best way to detect causes of unnecessary active scan in WLANs, where various independent and dependent parameters interact together. Both unsupervised and supervised methods are compared.

48  The Smart Cyber Ecosystem for Sustainable Development Data collected from real WLANs is used to train the ML algorithms. The authors deduced that a multilayer perceptron-based classifier model outperforms other models and accurately detect the cause of unnecessary active scanning. The work in [51] proposes an SDN-based framework for AP selection in WLANs, considering the QoS level required by users’ flows. The authors of [52] study the user association issue in an SDN–architecture. They developed heuristic algorithms that lead to high performance assuming unsaturated heterogeneous Markovian analytical model. The authors of [53] propose an admission control mechanism for VoIP calls in WLANs. A ML algorithm is used to predict the voice quality considering different parameters at the data link layer such as fraction of channel time used for video and normal traffic as well as estimated frame error rate for video and normal traffic. The authors of [54] leverage on the SDN paradigm to develop an algorithm that achieves effective distribution of traffic load in WLANs. The authors try to optimally distribute network resources and improve the overall performance.

2.7.2.2 Interference Mitigation Power control is a well-known approach used to mitigate interference in wireless networks. In SDN-based management and control of WLANs, the centralized CP can be used to implement the mechanisms for power control to minimize interference through coverage optimization of WLANs cells. Wireless interference classification is a process of identifying the type of wireless emitters exist in the local RF environment [55]. This is important for enabling coexistence of wireless technologies that operate in the same frequency band. ML-based solutions are being developed to achieve this goal. In [56], the authors propose a RL mechanism for interference mitigation in small cell networks. The algorithm represents the state of each AP as a binary variable that indicates whether the QoS requirement is violated. The action is a selection of power values from a set of power values. The reward is defined as the achieved rate. The algorithm iterates until a predefined level of QoS is met. The work in [57] develops a solution that uses ML-SVM for interference classification in wireless sensor networks from IEEE 802.11 signals and microwave ovens. Another deep learning approach for classification of WiFi, Zigbee, and Bluetooth was proposed in [58]. The authors defined fifteen classification tasks assuming a flat fading channel with additive white Gaussian noise. The research work of [59] compares different types of ML models for classifying signals, including deep feed-forward networks, deep convolutional networks, SVM and a multi-stage training algorithm. In [60], the authors propose a ML-based framework for mitigating the effect of jammers in WLANs, called “DeepWifi”. The system consists of an RF front end processing unit which applies a deep learning-based auto-encoder to extract spectrum-representative features. The system leverages the advances in ML algorithms to enhance the performance and security in WLANs. A deep neural network is then trained to classify signals as idle, WiFi, or jammer. In standard WiFi, the user backs off backs off regardless of the type of interference. However, DeepWiFi which is able to classify signals backs off when the interference is from another WLAN user, allowing user to operate in degraded mode and still receive non-zero throughput.

AI for Wireless Network Optimization  49

2.7.2.3 Channel Allocation and Channel Bonding Even with centralized control, optimal channel allocation problem in WLANs is difficult to be solved in an acceptable complexity level. Recently, researchers have been trying to leverage ML methods to find solutions in feasible time. In [61], the authors propose a ML method for assigning channels to WLANs APs. The method is based on passive monitoring of data in each cell. Using ML, it calculates the performance loss due to interfering users and finds the best channels for the cells that minimize interference. The algorithm minimizes airtime usage of interfering links in neighboring cells. Due to the dynamic nature of WLANs, the process is repeated iteratively. The authors of [62] concluded that a central control of APs is needed even if the network is influenced by neighboring unmanaged APs. Their approach results in a self-organizing system for channel allocation in WLANs based on cooperation between APs. The authors show that the proposed system leads to a stable network of high performance. In [63], the authors use ML techniques to learn implicit performance models from realworld measurements. The techniques do not need to know the details of interacting parameters. The authors used the developed model for channel allocation and power control.

2.7.2.4 Latency Estimation and Frame Length Selection Latency is a key factor that impacts the performance of modern mobile applications. In [64], the authors found that, latency depends on three main parameters: Channel utilization, the number of online devices, and the SNR. WiFi latency can be modeled using these related factors. The authors developed and compared the performance of supervised ML-based algorithms used to measure, characterize, and predict delay in large-scale WLANs. Training is implemented using data sets obtained from field measurements. Selecting a proper frame length is an important issue in WLANs, where it impacts the performance and users’ QoE in the network. The selection problem requires advanced techniques able to utilize information on practical settings in real-time. The work in [65] proposes an SDN-based solution for frame length selection in WLANs. The system proposes inclusion of ML techniques in SD-WLANs to optimize the selection of frame length for each user based on channel conditions as well as overall performance indicators. The supervised learning approach is used, where the algorithm is deployed on the management plane of the SDN architecture. The CP periodically feeds the algorithm with network knowledge about channel conditions and users’ state. The research work of [66] proposes a ML-based approach for the implementation of QoS management model in wireless networks. The ML system uses both supervised and unsupervised algorithms to identify key quality indicators for network users which represent an estimation of the quality as perceived by users considering influencing factors. Also, the ML concept is used for providing information about areas where corrective actions are required.

2.7.2.5 Handover Transparent handover with minimum overhead is still an open issue in WLANs. Though the 802.11r standard developed protocols that help implementing seamless handover between

50  The Smart Cyber Ecosystem for Sustainable Development WLAN cells, still APs and users need to be highly engaged in the handover process. This impacts the performance of APs, especially in dense deployments; wherein handover rate is expected to be high. With the advancement in AI and ML; coupled with the evolving SDN technology, researchers are trying to develop methods that allow low cost and successful transparent mobility among WLAN cells. An example effort is published in [67], where the authors developed an SDN-based solution for controlling and managing handover in WLANs. The proposed solution allows the devices to seamlessly move across cells without losing the QoS level. The researchers in [68] developed a framework to optimize the handover process and balance the network throughput and handover rate. Unsupervised ML algorithm is used to classify users according to their mobility patterns. Then, deep RL is used to optimize the handover process in each cluster. The received signal power by the user from APs is used as the state vector. The reward is considered to be the weighted sum between the handover rate and the throughput. In [67], the authors developed and tested an SDN-based solution for providing seamless handover in WLANs based on virtual APs. The solution maintains QoS requirements of real time applications in terms of packet loss and delay.

2.7.3 Cognitive Radio Networks Currently, ML-based spectrum sensing techniques are proposed. The existence of PUs is determined through two phases. In the first, signal features are extracted employing one of the spectrum sensing methods. In the second phase, decisions are made about PUs’ activity by applying ML algorithms. The authors of [68] used the energy detection approach to extract signal features which were used to train the K-mean-based ML model, whereas [69] proposes a probability vector as features of ML method. Probability vector has shown superior performance, whereby it alleviates the dimension of the feature vector and, hence, reduces the time duration of the ML model. In [70], the authors propose to use the eigenvalues/eigenvector as features, whereby these features are derived from constructing covariance matrix samples from different SUs’ signals. The authors of [71] derived the eigenvalues/ eigenvector features by applying principle component analysis of the signal samples. Based on information geometry theory and application, the work in [72] innovate novel features by measuring the distance between two probability distributions on a statistical manifold. The research efforts in [73–76] focus on finding the circularly characteristics that help in differentiating between the transmitted and noise signals. In their models published in [77, 78], the authors study the case where SUs catch more than one PU signals, whereas [79] investigated the mobility issue of CR systems.

References 1. Gacanin, H. and Ligata, A., Wi-fi self-organizing networks: Challenges and use cases. IEEE Commun. Mag., IEEE, 55, 7, 158–164, 2017. 2. Lohmüller, S., Cognitive Self-Organizing Network Management for Automated Configuration of Self-Optimization SON, in: PhD. Dissertation, University of Augsburg, 2019.

AI for Wireless Network Optimization  51 3. Thomas, R., DaSilva, L., MacKenzie, A., Cognitive Networks (book chapter), in: Cognitive Radio, Software Defined Radio, and Adaptive Wireless Systems, Springer, Germany, 2007. 4. Thang, V.V. and Pashchenko, F.F., Multistage System-Based Machine Learning Techniques for Intrusion Detection in WiFi Network. J. Comput. Networks and Commun., vol. 2019, Article ID 4708201, 13 pages, 2019. https://doi.org/10.1155/ 2019/4708201. 5. Mitchell, T., Machine Learning, McGraw Hill series in computer science, Mc- Graw-Hill, USA, 1997. 6. Liu, Y., Bi, S., Shi, Z., Hanzo, L., When Machine Learning Meets Big: A Wireless Communication Perspective. IEEE Veh. Technol. Mag., IEEE, 15, 63–72, 2020. 7. Kaufman, L. and Rousseeuw, P., Finding Groups in Data: An Introduction to Cluster Analysis, JohnWiley & Sons, US, 2018. 8. Andrews, J., Buzzi, S., Choi, W., Hanly, S., Lozano, A., Soong, A., Zhang, J., What will 5G be? IEEE J. Sel. Areas Commun., 32, 1065–1082, 2014. IEEE. 9. Agarwal, S., Kodialam, M., Lakshman, T., Traffic engineering in software defined networks. Proc. IEEE INFOCOM, pp. 2211–2219, 2013. 10. Mestres, A., Rodriguez-Natal, A., Carner, J., Barlet-Ros, P., Alarcón, E., Solé, M., Muntés-Mulero, V., Meyer, D., Barkai, S., Hibbett, M., Knowledge defined networking. ACM SIGCOMM Computer Communication Review, vol. 47, pp. 2–10, 2017. 11. Lim, S., Software Defined Network Detection System. Int. J. Recent Technol. Eng. (IJRTE), 8, 1391–1395, 2019. 12. Lin, P., Bi, J., Wolff, S., A west-east bridge based SDN inter-domain testbed. IEEE Commun. Mag., 53, 2, 190–197, 2015. 13. Xie, J., Yu, F., Huang, T., Xie, R., Liu, J., Wangz, C., Liu, Y., A Survey of Machine Learning Techniques Applied to Software Defined Networking (SDN): Research Issues and Challenges. IEEE Commun. Surv. Tutorials, 21, 393–430, 2019. 14. Elkhodr, M., Hassan, Q., Shahrestani, S., Networks of the Future: Architectures, Technologies, and Implementations, CRC Press Taylor & Francis Group, USA, 2018. 15. Kosmidesa, P., Adamopouloua, E., Demestichasa, K., Anagnostoua, M., Rouskasb, A., On Intelligent Base Station Activation for Next Generation Wireless Networks. The 6th International Conference on Emerging Ubiquitous Systems and Pervasive Networks, Elsevier, 2015. 16. Li, R., Zhao, Z., Chen, X., Zhang, H., Energy saving through a learning framework in greener cellular radio access networks. Proceedings of GLOBECOM, 1556–1561, 2012. 17. Yu, Y., Wang, T., Liew, S., Deep-Reinforcement Learning Multiple Access for Heterogeneous Wireless Networks. IEEE International Conference on Communications (ICC), IEEE, 2018. 18. Onireti, O., A Cell Outage Management Framework for Dense Heterogeneous Networks. IEEE Trans. Veh. Technol., 65, 2097–2113, 2016. 19. Mohammadi, M. and Al-Fuqaha, A., Enabling Cognitive Smart Cities Using Big Data and Machine Learning: Approaches and Challenges. IEEE Commun. Mag., 56, 94–101, 2018. 20. He, Y., Software-Defined Networks with Mobile Edge Computing and Caching for Smart Cities: A Big Data Deep Reinforcement Learning Approach. IEEE Commun. Mag., 55, 31–37, 2017.

52  The Smart Cyber Ecosystem for Sustainable Development 21. Jia, G., Yang, Z., Lam, H., Shi, J., Shikh-Bahaei, M., Channel assignment in uplink wireless communication using machine learning approach. arXiv preprint arXiv, 2001, 03952, 2020. 22. Zappone, A., Sanguinetti, L., Debbah, M., User association and load balancing for massive MIMO through deep learning. Proceedings of IEEE Asilomar Conference on Signals, Systems, and Computers, pp. 1262–1266, 2018. 23. Lin, P., Large-Scale and High-Dimensional Cell Outage Detection in 5G SelfOrganizing Networks. Proceedings of APSIPA Annual Summit and Conference, pp. 8–12, 2019. 24. Pervez, F., Jaber, M., Qadir, J., Younis, S., Imran, M., Fuzzy Q-learning-based usercentric backhaul-aware user cell association scheme. 2017 13th Interna­ tional Wireless Communications and Mobile Computing Conference (IWCMC), pp. 1840–1845, 2017. 25. Kumar, Y., Farooq, H., Imran, A., Fault Prediction and Reliability Analysis in a Real Cellular Network. 13th International Wireless Communications and Mobile Computing Conference (IWCMC), pp. 1090–1095, 2017. 26. Boutaba, R., Salahuddin, M., Limam, N., Ayoubi, S., Shahriar, N., Estrada-Solano, F., Caicedo, O., A comprehensive survey on machine learning for networking: evolution, applications and research opportunities. J. Internet Serv. Appl., Springer, 9, 16, 2018. https://doi.org/10.1186/s13174-018-0087-2 27. Song, R. and Willink, T., Machine Learning-Based Traffic Classification of Wireless Traffic. International Conference on Military Communications and Information Systems (ICMCIS), 2019. 28. Al-Issax, A., Bentaleb, A., Barakabitzex, A., Zinnery, T., Ghita, B., Bandwidth Prediction Schemes for Defining Bitrate Levels in SDN-enabled Adaptive Streaming. 15th International Conference on Network and Service Management (CNSM), 2019. 29. Fan, Z. and Liu, R., Investigation of machine learning based network traffic classification. Proceedings of ISWCS, pp. 1–6, 2017. 30. Song, C., Park, Y., Golani, K., Kim, Y., Bhatt, K., Goswami, K., Machine-learning based threat-aware system in software defined networks. Proceedings of IEEE ICCCN, pp. 1–9, 2017. 31. Glick, M. and Rastegarfar, H., Scheduling and control in hybrid data centers. Proceedings IEEE PHOSST’17, pp. 115–116, 2017. 32. Xiao, P., Qu, W., Qi, H., Xu, Y., Li, Z., An efficient elephant flow detection with cost-sensitive in SDN. 2015 1st International Conference on Industrial Networks and Intelligent Systems (INISCom), pp. 24–28, 2015. 33. Amaral, P., Dinis, J., Pinto, P., Bernardo, L., Tavares, J., Mamede, H., Machine learning in software defined networks: Data collection and traffic classification. Proceedings of IEEE ICNP’16, pp. 1–5, 2016. 34. Huang, T., Zhang, R., Zhou, C., Sun, L., QARC: video quality aware rate control for real-time video streaming based on deep reinforcement learning. ACM Multimedia Conference, ACM, 2018. 35. Azzouni, A. and Pujolle, G., Neutm: A neural network-based framework for traffic matrix prediction in SDN. IEEE/IFIP Network Operations and Management Symposium, 2018.

AI for Wireless Network Optimization  53 36. Jain, S., Khandelwal, M., Katkar, A., Nygate, J., Applying big data technologies to manage QoS in an SDN. Proceedings of IEEE CNSM’16, pp. 302–306, 2016. 37. Pasquini, R. and Stadler, R., Learning end-to-end application QoS from OpenFlow switch statistics. Proceedings of IEEE NETSOFT’17, pp. 1–9, 2017. 38. Letaifa, A., Adaptive QoE monitoring architecture in SDN networks: Video streaming services case. Proceedings of IEEE IWCMC’17, pp. 1383–1388, 2017. 39. Abar, T., Letaifa, A., Asmi, S., Machine learning based QoE prediction in SDN networks. Proceedings of IEEE IWCMC’17, pp. 1395–1400, 2017. 40. Comaneci, D. and Dobre, C., Securing Networks using SDN and Machine Learning. IEEE International Conference on Computational Science and Engineering, 2018. 41. Murudkar, C.V. and Gitlin, R.D., QoE-driven Anomaly Detection in Self Organizing Mobile Networks using Machine Learning. 2019 Wireless Telecommunications Symposium (WTS), pp. 1–5, April 2019. 42. Murudkar, C. and Gitlin, R., Machine Learning for QoE Prediction and Anomaly Detection in Self-Organizing Mobile Networking Systems. Int. J. Wireless Mobile Networks (IJWMN), 11, 2, April 2019. 43. Yao, H., Mai, T., Xu, X., Zhang, P., Li, M., Liu, Y., NetworkAI: An Intelligent Network Architecture for Self-Learning Control Strategies in Software Defined Networks. IEEE Internet Things J., 5, 4319–4327, 2018. 44. Zhu, L., Tang, X., Shen, M., Du, X., Guizani, M., Privacy-preserving DDoS attack detection using cross-domain traffic in software defined networks. IEEE J. Sel. Areas Commun., 36, 628–643, 2018. 45. Côté, D., Using machine learning in communication networks. J. Opt. Commun. Networks, 10, D100–D109, 2018. 46. Sultana, N., Chilamkurti, N., Peng, W., Alhadad, R., Survey on SDN based network intrusion detection system using machine learning approaches. Peer-Peer Network Appl., 12, 2, 493–501, 2019. 47. Moura, H., Alves, A., Borges, J., Macedo, D., Vieira, M., Ethanol: A Software-Defined Wireless Networking architecture for IEEE 802.11 networks. Comput. Commun., Elsevier, 149, 176–188, 2020. 48. Lei, T., Wen, X., Lu, Z., Li, Y., A semi-matching based load balancing scheme for dense IEEE 802.11 WLANs. IEEE Access, 5, 15332–15339, 2017. 49. Peng, M., He, G., Wang, L., Kai, C., AP Selection Scheme Based on Achievable Throughputs in SDN-Enabled WLANs. IEEE Access, IEEE, 7, 4763–4772, 2019. 50. Fulara, H., Singh, G., Jaisinghani, D., Maity, M., Chakraborty, T., Naik, V., Use of machine learning to detect causes of unnecessary active scanning in wifi networks. Proceedings of WoWMoM, pp. 1–9, 2019. 51. Ernst, J., Kremer, S., Rodrigues, J., A utility based access point selection method for IEEE 802.11 wireless networks with enhanced quality of experience. Proceedings of IEEE ICC, pp. 2363–2368, 2014. 52. Chen, J., Liu, B., Zhou, H., Yu, Q., Gui, L., Shen, X., QoS-driven efficient client association in high-density software-defined WLAN. IEEE Trans. Veh. Technol., 66, 7372– 7383, 2017. 53. Quer, G., Baldo, N., Zorzi, M., Cognitive call admission control for voip over ieee 802.11 using bayesian networks. In Proceedings of GLOBECOM, IEEE, pp. 1–6, 2011.

54  The Smart Cyber Ecosystem for Sustainable Development 54. Coronado, E., Villalon, J., Garrido, A., Wi-balance: SDN-based load-balancing in enterprise WLANs. IEEE Conference on Network Softwarization (NetSoft), pp. 1–2, 2017. 55. Jagannath, J., Polosky, N., Jagannath, A., Restuccia, F., Melodia, T., Machine learning for wireless communications in the internet of things: A comprehensive survey. Ad Hoc Networks, 93, 2019. Elsevier. https://doi.org/10.1016/j.adhoc.2019.101913. 56. Sanguanpuak, T., Guruacharya, S., Rajatheva, N., Bennis, M., Latva-Aho, M., Multioperator spectrum sharing for small cell networks: A matching game perspective. IEEE Trans. Wireless Commun., 16, 3761–3774, 2017. 57. Grimaldi, S., Mahmood, A., Gidlund, M., An SVM-based method for classification of external interference in industrial wireless sensor and actuator networks. J. Sens. Actuator Networks, 6, 9, 2017. https://doi.org/10.3390/jsan6020009 58. Kulin, M., Kazaz, T., Moerman, I., Poorter, E., End-to-end learning from spectrum data: a deep learning approach for wireless signal identification in spectrum monitoring applications. IEEE Access, 6, 18484–18501, 2018. 59. Youssef, K., Bouchard, L., Haigh, K., Krovi, H., Silovsky, J., Valk, C., Machine learning approach to RF transmitter identification. IEEE J. Radio Freq. Identif., 2, 197–205, 2018. 60. Davaslioglu, K., Soltani, S., Erpek, T., Sagduyu, Y., DeepWiFi: Cognitive WiFi with Deep Learning. IEEE Trans. Mobile Comput., 20, 429–444 2019. 61. Jeunen, O., Bosch, P., Herwegen, M., Doorselaer, K., Godman, N., Latre, S., A machine learning approach for ieee 802.11 channel allocation. 14th International Conference on Network and Service Management (CNSM), pp. 28–36, 2018. 62. Baid, A. and D. Raychaudhuri, D., Understanding channel selection dynamics in dense Wi-Fi networks. IEEE Commun. Mag., 53, 110–117, 2015. 63. Herzen, J., Lundgren, H., Hegde, N., Learning Wi-Fi Performance. 12th Annual International Conference on Sensing, Communication, and Networking (SECON), IEEE, 2015. 64. Sui, K., Zhou, M., Liu, D., Ma, M., Pei, D., Zhao, Y., Li, Z., Moscibroda, T., Characterizing and Improving WiFi Latency in Large-Scale Operational Networks. The 14th ACM International Conference on Mobile Systems, Applications, and Services, ACM, 2016. 65. Coronado, E., Thomas, A., Riggio, R., Adaptive ML-Based Frame Length Optimization in Enterprise SD-WLANs. J. Network Syst. Manage., Springer, 28, 850–881, 2020. 66. Ibarrola, E., Davis, M., Voisin, C., Close, C., Cristobo, L., QoE Enhancement in Next Generation Wireless Ecosystems: A Machine Learning Approach. IEEE Commun. Stand. Mag., 3, 63–70, 2019. 67. Košťál, K., Bencel, R., Ries, M., Trúchly, P., Kotuliak, I., High Performance SDN WLAN Architecture. Sensors, 29, 8, 1880, 8, 2019. 68. Wang, Z., Xu, Y., Li, L., Tian, H., Cui, S., Handover control in wireless systems via asynchronous multi-user deep reinforcement learning. IEEE Internet Things J., IEEE, 5, 4296–4307, 2018. 69. Sequeira, L., Cruz, J., Ruiz-Mas, J., Saldana, J., Fernandez-Navajas, J., Almodovar, J., Building an SDN enterprise WLAN based on virtual APs. IEEE Commun. Lett., 21, 374–377, 2017.

AI for Wireless Network Optimization  55 70. Kumar, V., Kandpal, D.C., Jain, M., Gangopadhyay, R., Debnath, S., K-mean clustering based cooperative spectrum sensing in generalized fading channels. Twenty Second National Conference on Communication (NCC), IEEE, pp. 1–5, 2016. 71. Lu, Y., Zhu, P., Wang, D., Fattouche, M., Machine learning techniques with probability vector for cooperative spectrum sensing in cognitive radio networks. IEEE Wireless Communications and Networking Conference, IEEE, pp. 1–6, 2016. 72. Sobabe, G., Song, Y., Bai, X., Guo, B., A cooperative spectrum sensing algorithm based on unsupervised learning. 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI), IEEE, pp. 1–6, 2017. 73. Wang, Y., Zhang, Y., Wan, P., Zhang, S., Yang, J., A spectrum sensing method based on empirical mode decomposition and k-means clustering algorithm. Wireless Commun. Mobile Comput., 2018, Article ID 6104502, 10, 2018. 74. Zhang, S., Wang, Y., Li, J., Wan, P., Zhang, Y., Li, N., A cooperative spectrum sensing method based on information geometry and fuzzy c-means clustering algorithm. EURASIP J. Wireless Commun. Networking, 17, 2019, 2019. 75. Hassan, Y., El-Tarhuni, M., Assaleh, K., Learning-based spectrum sensing for cognitive radio systems. J. Comput. Networks Commun., 13, 2012, 2012. 76. Ahmad, H., Ensemble classifier based spectrum sensing in cognitive radio networks. Wireless Commun. Mobile Comput., 2019, 16, 2019. 77. Huang, X., Hu, F., Wu, J., Chen, H., Wang, G., Jiang, T., Intelligent cooperative spectrum sensing via hierarchical dirichlet process in cognitive radio networks. IEEE J. Sel. Areas Commun., 33, 771–787, 2015. 78. Liu, B., Li, Z., Si, J., Zhou, F., Blind continuous hidden markov model-based spectrum sensing and recognition for primary user with multiple power levels. IET Commun., 9, 1396–1403, 2015. 79. Paul, A. and Maity, S., Kernel fuzzy c-means clustering on energy detection based cooperative spectrum sensing. Digital Commun. Networks, 2, 196–205, 2016. 80. Awe, O. and Lambotharan, S., Cooperative spectrum sensing in cognitive radio networks using multi-class support vector machine algorithms. 9th International Conference on Signal Processing and Communication Systems (ICSPCS), IEEE, pp. 1–7, 2015. 81. Xu, Y., Cheng, P., Chen, Z., Li, Y., Vucetic, B., Mobile collaborative spectrum sensing for heterogeneous networks: A bayesian machine learning approach. IEEE Trans. Signal Process., 66, 5634–5647, 2018.

3 An Overview on Internet of Things (IoT) Segments and Technologies Amarjit Singh

*

Jalandhar, Punjab, India

Abstract

The concept of IoT refers to the Internet of Things that can involve internet activity. But this can be done using the internetworking concept and aim to data information transfer. In other words, IoT can process for sharing information between virtual and system interaction. Using the IoT, it fetches the information using the sensors and other objects [1]. As one can with little of a stretch imagine, any certified duty to the improvement of the IoT ought to result from synergetic activities drove in different fields of data, for instance, communicate correspondences, informatics, contraptions, and human science. In such a capricious circumstance, it organizes this investigation to the people who need to advance toward this baffling train and add to its unforeseen development. It represents original dreams of this IoT perspective for enabling advances tested. What rises is that despite everything, significant issues will be confronted. This research paper includes the understanding of IoT and its different approaches. Keywords:  IoT, networking, sensors, wireless communication, sensor networks

3.1 Introduction Internet of Things (IoT) could be characterized assortment that is the internet; it is characterized as systems of systems that can associate millions of clients with a few typical internet conventions [2]. In IoT, the urban areas can be constructed where it manages with the parking spots, lighting, water system offices, commotion, and burn through, which can be checked continuously applications. We can fabricate keen homes that are extremely sheltered and progressively proficient to live. We can fabricate savvy conditions that can naturally be checking the contamination from air and water and empowering the early recognition of Tsunami, tremors, backwoods fires, and many annihilating debacles in the earth. A few modern, normalization, and study bodies engaged with the action of the development of answers to mollify the featured innovative prerequisites. This overview gives an image of the present cutting edge on the IoT [3]. IoT assumed expresses to a combine of extra problems about the framework’s body outlooks. Low resources will depict the things Email: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (57–68) © 2021 Scrivener Publishing LLC

57

58  The Smart Cyber Ecosystem for Sustainable Development framing the IoT to the amount of together estimation and essential limit. The future progressions of action need to give outstanding assumed to affirm adequacy additional than the understandable adaptability issues (Figures 3.1 and 3.2). The IoT area prompts the universe of innovation and correspondence to another period where items can impart, register, and change the data according to the prerequisites. The IoT is a modification in outlook in the IT sector [4]. As from definition, the internet is the PC that organizes all-inclusive interconnected and uses the standard convention for serving the millions of clients everywhere the world. The guideline nature of the IoT, however, is the high impact; it will have on a couple of parts of normal day-by-day presence and lead of possible customers. From a private customer, the clearest impacts of the IoT would be perceptible in together the occupied and private arenas. Devilfish lived, e-prosperity, improved learning is only two possible application circumstances in which the new perspective will expect the significant activity in a matter of seconds. So, likewise, from business customers, the clearest outcomes will be comparably perceptible in fields, for example, motorization and present-day, co-appointments, business/process of the officials, and brilliant conveyance of people and items [5]. IoT develops quickly; it will be proceeding with by 2025, and there is the overall pattern utilizing the IoT. In IoT, information is gathered from different sensors, transmitted over remote systems, and afterward investigated. The detected and investigated information will

IoT Internet

Mobile Network

IoT Communication

Figure 3.1  Process of IoT communication.

IoT Communication

Figure 3.2  IoT communication devices.

IoT Segments and Technologies  59 be used to control actuators. IoT permits articles to be detected and controlled remotely or basically across existing system foundation, making open doors for more straightforward coordination between the physical world and PC-based frameworks, and bringing about improved proficiency, precision, and financial advantage [6]. IoT sense can allude to a wide assortment of gadgets, for example, heart checking inserts, biochip transponders on livestock, electric shellfishes in seaside waters, vehicles with worked in sensors, and DNA examination gadgets for natural/food/pathogen observing or field activity gadgets that help firemen in search and salvage tasks. These gadgets gather valuable information with the assistance of different existing innovations and afterward independently stream the information between different gadgets.

3.2 Features of IoT Main features of the IoT include sensors, connectivity among various nodes, artificial intelligence, and smart devices. Some of the key features exist and are mentioned below: 1. Sensors: IoT uses various kinds of sensors which can get the information or data from numbers of nodes, connected with specific networks. This would get the data gathering between sensors devices and predict and addresses various system equipment [7]. Sensors are also reduced the human workload by collecting the information. 2. Connectivity: It enables the new networking devices connectivity process with a combination of IoT. So, networks would be acquiring the information whether it is based on a small or large network. IoT communicates with networks and sends the data. But connectivity is an important factor to utilize any IoT device. 3. Artificial Intelligence: IoT makes any process to be virtually and smart which would enhance the performance. Data collection using an artificial intelligence algorithm is efficient [8]. This technique is a smart process can utilize according to user need and trend concept. 4. Smart Devices: IoT uses various types of sensors to resolve the purpose of user requirements. These devices depend upon their need and cost constraints [9]. According to the IoT, the process uses smaller and cheaper and more powerful devices which would be beneficial for scalability and reliability.

3.3 IoT Sensor Devices • Temperature Sensor A temperature sensor facilitates and measures the temperature and changes over it into an electrical sign. They have a significant job in the environment, agriculture, and industries. For instance, these sensors can recognize the temperature of the dirt, which is progressively useful in the creation of harvests. There are numerous kinds of temperature sensors and proficient, simple to introduce, and solid that reacts to human action [10]. The sensors take a shot

60  The Smart Cyber Ecosystem for Sustainable Development

•

•

•

•

•

at the connection between the metals and the temperature, as the opposition of the gadget is legitimately relative to the temperature. The broadly utilized DHT is the temperature and dampness sensor, which is the fundamental, minimal effort, advanced, and capacitive sensor. Pressure Sensor A weight sensor detects the weight applied, i.e., power per unit zone, and it changes over into an electrical sign. It has high significance in climate gauging. There are different Pressure sensors accessible in the market for some reason. For instance, if there are any water spills in the private or business regions, a weight sensor should be introduced to check if there are any breaks and measures the weight [11]. Another, e.g., all the cell phones, wearables have these barometric weight sensors incorporated into them. Smoke Sensor A smoke sensor recognizes smoke and its degree of accomplishment. These days, the producers of the sensor actualize it with a voice alert additionally informs in our cell phones. The smoke sensor is of two kinds: the optical smoke sensor and the ionization smoke sensor [12]. The optical smoke sensor likewise called photoelectric smoke alerts works utilizing the light dispersing guideline. The caution contains a beat infrared LED which beats a light emission into the sensor chamber to check for smoke particles. Proximity Sensor A proximity sensor is a sensor ready to perceive the nearness of close-by objects with no physical contact. A nearness sensor regularly discharges an electromagnetic field or a light emission radiation and searches for varieties in the field or brings a signal back. The most regular use of this sensor is utilized in vehicles [13]. While you are taking the opposite, it identifies the items or obstructions and you will be frightened. Likewise, it is utilized in retails, exhibition halls, stopping in air terminals, shopping centers, and so forth. Inductive, capacitive, photoelectric, and ultrasonic are the sorts of a closeness sensor. The inductive sensors identify the metal objective while the photoelectric and capacitive sensors distinguish the plastic and natural targets. In cell phones, it detects the client’s face is close to the telephone during a call. Accelerometer Sensor Accelerometers in cell phones are utilized to distinguish the direction of the telephone. The gyroscope adds measurement to the data provided by the accelerometer by following revolution or curve [14]. A spinner has three gyroscopic sensors mounted symmetrically. Accelerometers and gyrators are the sensors of decision for obtaining increasing speed and rotational data in rambles, PDAs, vehicles, planes, and versatile IoT gadgets. IR Sensor IR sensor is an automated gadget, which detects certain qualities of its environmental factors by producing infrared radiation. It can gauge the warmth being radiated by an article and quantifies the separation [15]. It has been executed in different applications. It is utilized in radiation thermometers to rely upon the material of the article. IR sensors are additionally utilized in flame screens and dampness examination. IR sensors are utilized in gas

IoT Segments and Technologies  61

•

•

analysers that utilization the change attributes of gases in the IR area. Two sorts of strategies are utilized to gauge the thickness of gas, for example, dispersive and nondispersive. IR imaging gadgets are utilized for warm imagers and night vision [16]. Optical Sensor The optical sensors convert light beams into an electronic sign; it gauges a physical amount of light and changes into a coherent structure, possibly advanced structure. It recognizes the electromagnetic vitality and sends the outcomes to the units [17]. It includes no optical filaments. It is an extraordinary aid to the cameras on cell phones. Additionally, it is utilized in mining, substance industrial facilities, processing plants, and so forth. LASER and LED are the two distinct kinds of the light source. Optical sensors are fundamental pieces of numerous regular gadgets, including PCs, copiers, and light apparatuses that turn on consequently in obscurity. What is more, a portion of the basic applications incorporates alert frameworks, coordinated for photographic flashes, and frameworks that can recognize the nearness of articles [18]. Gas Sensor A gas sensor or a gas identifier is a gadget that distinguishes the gas in a region, which is extremely useful in security frameworks. It is a rule, recognized a gas spill in a zone, that outcomes are sent to a control framework or a small-scale controller, that at long last closes down [19]. It can identify ignitable, combustible, and harmful gases.

3.4 IoT Architecture The IoT architecture varies from their functional and their resolution. IoT engineering innovation principally comprises of four significant parts (Figure 3.3): • Sensor: Sensors are the contraptions that can verbalize, recognize, and process data over the framework. They may relate these sensors either through wired or remote. A large portion of the sensors needs availability through the sensor’s entry ways [20]. The association of sensors can be through a Local Area Network (LAN). • Gateway and Networks: A huge amount of information is delivered by this sensor and need fast entry ways and networks to interchange the material. This system can be type of LAN (Wi-Fi, Ethernet, etc.). • Cloud Service: Cloud goes under the management services which create the information through assessment, the leading group of the device, and security controls. Other than these safety controls and devices, the officials who use the cloud move the data to the end customer’s application, for instance, retail, healthcare, emergency, and energy. • Application Layer: The application layer actualizes the working of IoT. For this, an application is required with the relating gadget to finish the necessary assignment.

62  The Smart Cyber Ecosystem for Sustainable Development

Sensors/Devices

Gateways and Networks

Cloud Service

Application Layer

Figure 3.3  IoT architecture.

3.5 Challenges and Issues in IoT IoT has enough challenges and resolutions according to its behavior approach. In this section, our main approach is to understand the issues and use of the IoT accordingly. • Lack of Outdated Devices The use of outdated devices is the major factor to lack the security in IoT [21]. There should be advanced technology so that devices can communicate properly and well known about the recent threats. • Connectivity Issue This is, perhaps, the most disregarded test since the information network has immensely improved. However, there still exist a few zones where information availability is an IoT usage challenge. It includes how IoT gadgets converse with the passage and the cloud and what information position does they create. Most IoT doors accessible are good with GPRS and Wi-Fi/LAN. In this way, the need is for a reasonable edge layer that interprets transport and information group conventions to send information to the IoT stage [22]. Defining the precise blend of these resolutions before ongoing with an IoT implementation will help in going far. • Information Security Issues Clients are uncertain about information security. There is additionally an opportunity of corporate secret activities to increase licensed innovation [23]. Thus, IoT specialist co-ops need to guarantee that their information

IoT Segments and Technologies  63

• •

•

•

•

•

will be protected. It can take this information care by utilizing an extensive administration mode, which gives secure access to touch reports and information. This period of planning that defines different information related to security arrangements is critical for effective IoT execution. Vulnerabilities Ordinary hacking of prominent targets keeps this threat continually in the rear of our psyches. The results of treachery and disavowal of administration could be undeniably more genuine than a trade-off of protection [24]. Changing the blend proportion of disinfectants at a water treatment plant or halting the cooling framework at an atomic force plant may put an entire city in impending peril. Scalability This is not quite a bit of an issue; however, they bound it to turn into an issue essentially comparable to nonexclusive buyer cloud as the number of gadgets inactivity rises [25]. This will build the information limit required, and the time required for confirming exchanges. Limited Sensor Availability Principal sensor types, for example, temperature, light, movement, and laser scanner, are effectively accessible. Ongoing advances in microelectronics, combined with progress in strong state sensors, will make the uncovered sensors less of an issue [26]. The test will be in making them all the more separating in swarmed, uproarious, and increasingly complex situations. The use of calculations like fluffy rationale vows to make this less of an issue. Analytics Challenges We comprehend the certified estimation of an IoT plan through huge encounters got from the assembled IoT data [27]. This demands an unrivalled assessment stage fit for dealing with the proportion of data to be added to the game plan at a later point. Data analytics accessories need to recall this while thinking up the IoT execution building to incorporate data planning, decontaminating, and depiction. Accordingly, enough space for extensibility to incorporate a nonstop or insightful assessment to an IoT game plan can help disentangle this fundamental IoT execution challenge. Usage of Energy-Based Equipment Energy utilization is a significant issue in IoT giving an energy source they consider to different savvy questions and issue. Remote force innovations are an intriguing exploration region wherein force can be transmitted at some separation to the savvy objects [28]. Another exploration territory is a directing improvement in WSN to decrease energy utilization by choosing the best course.

3.6 Future Opportunities in IoT • Increased Usage of IoT We expect the Web of Things gadgets, for example, machines and sensors to expand their use in 2025. IoT will develop at a compound yearly development

64  The Smart Cyber Ecosystem for Sustainable Development

•

•

•

•

pace of 22.5% from 2022 to 2025. IoT web innovation is the following significant advance in making the world an associated place. 6G IoT Need 6G is energetic to the IoT or single scheme for millions of consumptions. In 10 years, time period from 2032 to 2040, IoT devices will grow over 160 billion, and the development from 5G to 6G in develop IoT is substantial. With a 5G organize, a solitary cell can deal with, up to one million gadgets. Increased Client Adoption Next 10 years, you will see an immense change in IoT when the move away will occur from the consumer-based IoT like the market lemon of robotics. The subsidizing development of the shopper-based IoT will diminish, and the future will be the year for the mechanical IoT foundation and stage [29]. This IoT trends will expect time to develop. Smart Homes Concepts Before, we have seen the IoT applications have flooded with brilliant home innovation, and this will have proceeded in the blink of an eye so the home can get progressively intelligent [30]. Individuals would not immediate the gadgets; rather, the gadgets will mention to the individuals of the house what they ought to do. Health Industry Grow IoT Gadgets are being used by the medicinal services segments, and these segments will confront a consistent yet stable turn of events. Medical gadgets use the Cloud and put away the pictures for Intelligent Systems [31]. This will be something to note down that along these lines, the legislature will receive the best advantage in return [32]. We exceptionally empower the IoT trends in the healthcare world is high [33].

3.7 Discussion New vulnerabilities, for example, unbound correspondence channels, malignant exercises in the system, and unbound physical gadgets, acquaint new kinds of dangers with the IoT systems [34]. This additionally confirms that IoT gadgets are the objectives of surface assaults because of their unpredictable fixing and updates: regularly, the gadgets accompany negligibly or possibly no confirmation or encryption by any means [35]. For the most part, these gadgets are sent in an antagonistic situation and accessible consistently; thus, there might be negligible assurance any unlawful physical access. Affirmation and encryption may be convincing courses of action in lightening security issues in IoT [36]. For low-power, computationally, and resource obliged devices, the utilization of suitable affirmation and encryption is still in its most punctual stages and does not guarantee the evasion of noxious center points in the framework, for instance, contaminated contraptions or machines. In view of its solace, producers apply hard-coded accreditations or passwords. It normally prompts a critical confirmation disappointment. From this overview, it is seen that ebb and flow research on gadgets’ security has for the most part focussed on improving lightweight confirmation and encryption for low-force and asset compelled gadgets.

IoT Segments and Technologies  65 On the other hand, ensuring about the guiding show at the framework layer and executing trust and reputation-based malicious center point area bears an all the way delay, correspondence overhead, and a high counterfeit positive rate. The disclosures from this examination show that although check alone may not be satisfactory for IoT security, the example of IoT security instruments is to go after lightweight, normal, and multifaceted affirmation, especially at the framework and application layers. To facilitate contraptions’ security issues, lightweight and insignificant exertion encryption are proposed for the physical layer. Considering, as appeared by the IoT security structure, security balance remembers all the layers for the key IoT building, explicitly, wisdom, framework, and application, although it is seen that most by far of the current segments apply to the framework layer. It also can be gathered that a fitting IoT peril showing might be important in arranging fascinating IoT security control.

3.8 Conclusion Web of Things depends upon the Internet and sensors advancement which make correspondence possible among devices by realizing original shows. After doing the composing study, we talked about some critical issues like the meddled with accessibility among contraptions affecting the correspondence. In like manner, there is a comparability issue with devices. The security of contraptions during the correspondence technique and security of the correspondence channels or associations is in like manner a critical issue. Lots of work is to be practiced for the headway and progress of this field; still, there is more work to do, more standardization of development, shows, and hardware is required to make the reliable and secure space of the IoT. What is to come is depending upon the IoT, so a lot of exercises at the execution level.

References 1. Al-Fuqaha, A., Guizani, M., Mohammadi, M., Aledhari, M., Ayyash, M., Internet of things: A survey on enabling technologies, protocols, and applications. IEEE Commun. Surv. Tutorials, 17, 2347–2376, 2015. 2. Welbourne, E., Battle, L., Cole, G., Gould, K., Rector, K., Raymer, S. et al., Building the internet of things using rfid the rfid ecosystem experience. IEEE Internet Comput., 13, 48–55, 2009. 3. Chen, H.C., Faruque, M.A.A., Chou, P.H., Security and privacy challenges in IoT-based machine-to-machine collaborative scenarios. 2016 International Conference on Hardware/ Software Codesign and System Synthesis (CODES+ISSS), Pittsburgh, PA, pp. 1–2, 2016. 4. Miao, W., Ting, L., Fei, L., ling, S., Hui, D., Research on the architecture of Internet of things. IEEE International Conference on Advanced Computer Theory and Engineering (ICACTE), Sichuan province, China, pp. 484–487, 2010. 5. Parashar, R. and Abid Khan, N., A survey: the Internet of Things. Int. J. Tech. Res. Appl., 4, 3, 251–257, May-June, 2016. 6. Li, W., Song, H., Zeng, F., Policy based Secure and Trustworthy Sensing for Internet of Things in Smart Cities. IEEE Internet Things J., 2017. 7. Sha, K., Wei, W., Andrew Yang, T., Wang, Z., Shi, W., On security challenges and open issues in Internet of Things. Futur. Gener. Comput. Syst., 83, 326–337, 2018.

66  The Smart Cyber Ecosystem for Sustainable Development 8. Chernyshev, M., Baig, Z., Bello, O., Zeadally, S., IEEE Internet of Things (IoT): Research. 5, 3, 1637–1647, 2018. 9. RiahiSfar, A., Natalizio, E., Challal, Y., Chtourou, Z., A roadmap for security challenges in the Internet of Things. Digit. Commun. Networks, 4, 2, 118–137, 2018. 10. Srinivas, J., Mukhopadhyay, S., Mishra, D., Ad Hoc Networks Secure and efficient user authentication scheme for multi gateway wireless sensor networks. Ad Hoc Networks, 54, 147–169, 2017. 11. Giuliano, R., Mazzenga, F., Neri, A., Vegni, A.M., Member, S., Security Access Protocols in IoT Capillary Networks, in: IEEE Internet of Things Journal, 4, 3, 645–657, 2017. 12. Elgenaidi, W., Rao, M., Dooly, G., Toal, D., Sensors and Actuators A: Physical A secure end-toend IoT solution. Sensors Actuators A. Phys., 263, 291–299, 2017. 13. Choi, S., Yang, C., Kwak, J., System Hardening and Security Monitoring for IoT Devices to Mitigate IoT Security Vulnerabilities and Threats. 12, 2, 906–918, 2018. 14. Singh, S., Advanced lightweight encryption algorithms for IoT devices: survey, challenges and solutions. J. Ambient Intell. Humaniz. Comput., 0, 0, 0, 2017. 15. Shin, D., Sharma, V., Kim, J., Kwon, S., You, I., Secure and Efficient Protocol for Route Optimization in PMIPv6-Based Smart Home IoT Networks. IEEE Access, 5, 11100–11117, 2017. 16. Shin, D., Sharma, V., Kim, J., Kwon, S., You, I., Member, S., Secure and Efficient Protocol for Route Optimization in PMIPv6 - Based Smart Home IoT Networks, in: IEEE Access, 5, pp. 11100–11117, 2017. 17. Agrawal, S. and Vieira, D., A survey on Internet of Things. Abakós, Belo Horizonte, 1, 2, 78 – 95, maio 2013. 18. Muriel, D. and Juan, F., Expanding the learning environment: combining physicality and virtuality The Internet of Things for eLearning. IEEE International Conference on Advanced Learning Technologies (ICALT), Sousse, Tunisia, pp. 730–731, 2010. 19. Pande, P. and Padwalkar, A.R., Internet of Things –A Future of Internet: A Survey. Int. J. Adv. Res. Comput. Sci. Manage. Stud. Res. Article / Paper / Case Study, 2, 2, 354 – 361, February 2014. 20. Riazul Islam, S., Kwak, D., Humaun Kabir, M., Hossain, M., Kwak, K.S., The internet of things for healthcare: a comprehensive survey. IEEE Access, 3, 678–708, 2015. 21. Zhang, H. and Zhu, L., Internet of Things: Key technology, architecture and challenging problems. 2011 IEEE International Conference on Computer Science and Automation Engineering, Shanghai, pp. 507–512, 2011. 22. Rahul, A., Gokul Krishnan, G., Unni Krishnan, H., Rao, S., Near Field Communication (NFC) Technology: A Survey. Int. J. Cybern. Inf. (IJCI), 4, 2, pp. 251–257, April 2015. 23. Usha Devi, Y. and Rukmini, Dr. M.S.S., IoT in Connected Vehicles: Challenges and Issues- A Review. International Conference on Signal Processing, Communication, Power and Embedded System, 2016. 24. Van Kranenburg, R., A critique of ambient technology and the seeing network of RFID, 2008. http://www.networkcultures.org/_uploads/notebook2_theinternetofthings.pdf 25. Khan, R., Khan, S.U., Zaheer, R., Khan, S., Future internet: the internet of things architecture, possible applications and key challenges. 2012 10th International Conference on Frontiers of Information Technology, pp. 257–260, 2012. 26. Al-Fuqaha, A., Guizani, M., Mohammadi, M., Aledhari, M., Ayyash, M., Internet of things: A survey on enabling technologies, protocols, and applications. IEEE Commun. Surv. Tutorials, 17, 2347–2376, 2015. 27. Zhu, C., Leung, V.C., Shu, L., Ngai, E.C.H., Green internet of things for smart world. IEEE Access, 3, 2151–2162, 2015.

IoT Segments and Technologies  67 28. Bilal, M., A Review of Internet of Things Architecture, Technologies and Analysis Smartphone based attacks against 3d printers. arXiv preprint arXiv, 1708, 04560, 2017. 29. Borkar, S. and Pande, H., Applications of 5G next generation network to internet of things. Internet of things and Applications, International Conference on 2016, pp. 443–447. 30. Kyriazis, D., Varvarigou, T., Rossi, A., White, D., Cooper, J., Sustainable smart city IoT applications: heat and electricity management & Eco- conscious cruise control for public transportation. IEEE 14th International Symposium and Workshops on World of Wireless, Mobile and Multimedia Networks, pp. 1–5, 2013. 31. Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M., Internet of Things (IoT): A vision, architectural elements, and future directions. Future Gener. Comput. Syst., 29, 1645–1660, 2013. 32. Bassi, A. and Horn, G., Internet of Things in 2020: A Roadmap for the Future. Eur. Commission: Inf. Soc. Media, 22, 97–114, 2008. 33. Alamri, A., Ansari, W.S., Hassan, M.M., Hossain, M.S., Alelaiwi, A., Hossain, M.A., A survey on sensor-Cloud: architecture, applications, and approaches. Int. J. Distrib. Sens. Netw., 1–18, 2013, 2013. 34. Shrouf, F., Ordieres, J., Miragliotta, G., Smart factories in Industry 4.0: A review of the concept and of energy management approached in production based on the Internet of Things paradigm. Proceedings of the 2014 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), Selangor Darul Ehsan, Malaysia, pp. 697–701, 9–12 December 2014. 35. Connolly, D., Lund, H., Mathiesen, B., Smart Energy Europe: The technical and economic impact of one potential 100% renewable energy scenario for the European Union. Renew. Sustain. Energy Rev., 60, 1634–1653, 2016. 36. Motlagh, N.H., Khajavi, S.H., Jaribion, A., Holmstrom, J., An IoT-based automation system for older homes: A use case for lighting system. Proceedings of the 2018 IEEE 11th Conference on Service-Oriented Computing and Applications (SOCA), Paris, France, pp. 1–6, 19–22 November 2018.

4 The Technological Shift: AI in Big Data and IoT Deepti Sharma1*, Amandeep Singh2 and Sanyam Singhal3 CSE Department, KCC Institute of Legal and Higher Education, Greater Noida, India Department of Internet Working, Dalhousie University, Halifax, Nova Scotia, Canada 3 Department of Physics, Indian Institute of Technology Bombay, Mumbai, India

1 2

Abstract

Changes have always been a major part of life; we experience changes in our body, thoughts, surroundings, and so do with the technology. Artificial Intelligence (AI)–based products brought a revolution in the modern world, making a global impact on technology. This technology not only gave life to a machine but also imparted emotions into it. Whereas, when AI connects with Internet of Things (IoT), it enabled us to operate the machines remotely. During the entire communication process, a huge volume of data chunks is transferred to the cloud so that machines can communicate more efficiently. In this chapter, we will discuss the present scenario of AI in Big Data and the IoT. The languages are utilized in NLP and ANN and their algorithms to predict the best possible results in an optimized manner. We look deeper into IoT modifications which will enhance the properties of the system and its contribution in longer productivity. The major objective of this chapter is to dig deeper into a broad range of applications which can be consumed by AI and ML technology, outcomes of these modifications by keeping economic factors into account, and to have a predictive analysis of the AI systems. Keywords:  Big data, machine learning, artificial intelligence, data set, deep learning

4.1 Introduction Artificial Intelligence (AI) is the expression of human intelligence processed by computer systems. These processes include learning and reasoning with self-correction. AI systems generally show some behavior’s like human intelligence that includes planning, learning, reasoning, problem-solving, knowledge presentation, perception, and creativity [1, 2]. AI is present everywhere in today’s modern time and will spread its roots further in almost every sector in the future. Some prevalent examples of AI can be seen in websites or mobile applications like Amazon and Flipkart; what they do is they actually keep a record of what the user has searched before and then when the user uses the application for the next time, it automatically suggests its user what one can think of buying. Moreover, further new chatbots are also available in the market which helps people to do things for them like AmazonAlexa and Google Home; they are found to be very useful and productive for senior citizens *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (69–90) © 2021 Scrivener Publishing LLC

69

70  The Smart Cyber Ecosystem for Sustainable Development and differently abled people who can command to turn on/off the appliances for them even it is observed that people are trying to ignore their loneliness with them [3]. A feature is now readily available in Google photos and in Apple iPhones to distinguish different people based on their faces and helps the user to see a segregated collection of photos of a particular person. Last but not the least, its applications can be seen in credit card fraud detection and in spotting spam messages/calls [4]. AI is a very different approach to do some particular kind of work but with the help of systems. When AI is unified with Machine Learning (ML), where a system can feed large volumes of data, it then utilizes to learn the technique to carry out the specific task. For example, unified applications of ML and AI are being used in many neuro- and cardiovascular surgeries [5]. Big Data is the raw data available in large volumes for further usage in different areas. Usually, we apply ML algorithms on Big Data to predict the outcomes and train a network in an optimized way [6]. For example, mining patient’s medical records can be used to evaluate effectiveness of several recovery options and to discover the most effective treatment to fight various diseases. However, this data is highly confidential and cannot be disclosed to everyone or kept on a public platform. Similarly, the applications monitoring such data through mobile phones [7] should also follow some protocols for hiding data from the public and could be merged with ML and AI to get a forecast about what can happen in upcoming years. AI needs data to build a more intelligent system. For example, a ML face recognition application which is quite prominent in today’s world takes a set of around 1000 or more training sets to get trained and raise its efficiency. Morrison quoted that “The data you start with is Big Data, but to train the model, that data needs to be structured and integrated well enough that machines are able to reliably identify useful patterns in the data” [8]. The data which is used in AI and ML is already clean and do not require any sort of modifications; duplicate elements are already removed from the data and all other unnecessary things are popped off at the initial step. Then, the necessary ML algorithm is selected (according to the purpose) and is applied on the selected input training data set. After going through several phases and fulfilling numerous conditions, the output is generated, and further, Big Data analysis is applied on it to get optimized results which can increase the efficiency of the system and could be trusted to study the patterns in future [9]. As shown in Figure 4.1, Big Data plays a major role in making a successful AI and ML-enabled application since both of them require a good volume of data which can be readily used to study the trends and shifts in the records which can help the machine to get trained well and learn the paradigms occurred and to give us optimized results. So far, we have seen how we can make our system learn by providing it a vast data set but the problem which is persisting with the system is that the machine cannot be operated

Data segregation Removal of unnecessary data STEP 1

Application of Al, Utilisation of ML Algorithms STEP 2

Figure 4.1  Steps involved in using AI and Big Data together.

Big Data Analytics is applied and output is compared with expected output STEP 3

AI in Big Data and IoT  71 from anywhere, anytime. To remove this issue, we need to merge the AI and Big Data with Internet of Things (IoT) technology. IoT gives flexibility to the user to do anything from anywhere at any time; this makes a machine more intelligent that it is now capable of doing work on its own. IoT involves a wide range of sensors which helps the machine to collect data from different time intervals, and hence, a huge volume of data is collected and provided to the machine as a training set [10–12]. In this chapter, we are going to focus upon the technological changes which are introduced by AI, Big Data, and IoT. How technology has shifted to these new techniques, why are they prominent in the present world, and who they can be further molded with other several techniques to provide more ease and comfort to mankind.

4.2 Artificial Intelligence Automating the tasks along with the thoughts done by humans by the usage of computational and robotic facilities is what succinctly defines AI. The idea of machines eventually replicating or even surpassing human capabilities in various aspects has drawn all kinds of attention. Big Data and the world of the Internet are not untouched by this idea. The field of Big Data analysis mandates automation due to enormous volumes of data, ruling out manual statistical interpretation. AI systems have become an integral part of evaluating Big Data because of two major reasons: increased computational power by CPU-GPU parallelism [13] and efficient algorithms. CPU-GPU parallelism has enabled implementation of Big Data analysis beyond the limited availability of Super Computing facilities. The field has widened to individual researchers and analysts. This widespread involvement has also heavily gained from rising popularity and usage of the concepts of AI. We use the concepts of volume, velocity, and variety to describe Big Data [14]. Volume is used to define the amount of data available; variety describes the categorical variations; and velocity conveys how fast our data is growing and updating. AI systems fundamentally “learn” from the volume through various algorithms, validate the accuracy of learning on various varieties, and update their performance with newer data. Exponential rise of total data on the Internet has made it essential for humans to consider computational automation to make sense of this data labyrinth. High computational power, diverse programming packages, and fast internet to use stateof-the-art servers and cloud storage facilities have strongly motivated the linkage of AI and Big Data analytics. It becomes evident from the fact that the idea of AI emerged nearly 7 decades ago; however, it gained prominence in the last two decades, owing to the enabling of practical usage. Two major subdivisions of AI that work towards deriving meaning out of Big Data are Deep Learning and IoT.

4.2.1 Machine Learning At the very core, ML leverages various statistical methods for finding the best possible estimating functions or graphs [15] by minimizing a cost function (a measure of deviation) [16] or maximizing a utility function (a method to incentivize the algorithm on accuracy) [17]. These estimated functions or graphs are then utilized to make numerical or categorical predictions.

72  The Smart Cyber Ecosystem for Sustainable Development These predictions, dependent on learning from large sets of data, have now become a routine in the modern world. Some of the prominent examples are cancer detection [18], malicious mobile applications detection [19], and high-frequency trading [20]. All these fields are dependent on Big Data and their heavy computation demands automation, hence the usage of IoT. On a high level, any ML system is created using following steps: 1. Data Preprocessing: Big Data floating on the web is crude and complex and it requires parsing. It is present in various formats like SDF, CSV, JSON, and JPG. Big Data is collected and processed using modules like MapReduce [21]. Data outliers are removed; data features are scaled to remove apparent bias for the learning algorithm. This cleaning is essential to get the best possible accuracy. For computational ease, non-Gaussian data distribution is approximated to a Gaussian distribution via various transformations [22]. 2. Data Imputation: On many occasions, raw data instances have missing values for some fields. These missing values are filled by modeling with non-empty fields of other data instances using various statistical methods like mean imputation, stochastic regression imputation, and, a more recent method, probabilistic principal component analysis [23]. 3. Dimensionality Reduction: Often, the data has too many fields per instance which lead to exponential or polynomial growth in computational complexity due to the fact that most of the ML algorithms are non-linear. Dimensionality reduction means reduction of data from a high-dimensional manifold to a low manifold through various mapping techniques [24, 25]. It represents a trade-off between accuracy and computational speed. Hence, for Big Data, where the amount of data is enormous, we often use dimensionality reduction. It is commonly used for research and application in image analysis and computer vision [26], where a lossy image does not lead to a significant drop in statistically important data. 4. Model Selection: A training model or algorithm is selected to train the system about the data and make predictions. We select the algorithm because of the following: 1) nature of task, i.e., regression [27] or classification [28]; 2) size of dataset; and 3) sensitivity and specificity trade-off. These factors prominently decide the computation duration and accuracy of the model. Often, a collection of models called ensemble is utilized to increase accuracy where there is scope for increased computation (small datasets), or accuracy is essential and computation speed can be traded off (cancer detection). 5. Cross-Validation and Hyperparameter Tuning: Cross-validation refers to testing the accuracy of the model on a portion of the training segment of the dataset before applying the model on test data. The portion of training data that is used for cross-validation has not been seen by the algorithm during training. This analysis helps to avoid underfitting or overfitting of the model. Hyperparameters which determine the mathematical nature of the prediction model are tweaked and various iterations are done to find the best set of hyperparameters for a given number of iterations. The models that have this facility are often the best estimators due to their flexibility.

AI in Big Data and IoT  73 6. Testing the Model: For the final stage of building an IoT model, it is tested on test data and if any aberrations are found; steps 1–5 are repeated. As mentioned before, IoT has been widely used in automation of alphanumeric character recognition; cancer detection by formulating it as a classification problem on the basis of biological parameters as features, image detection, and computer vision involves usage of geometry (like epipolar geometry) along with classification algorithms to formulate a visual problem into an algorithmic one.

4.2.2 Further Development in the Domain of Artificial Intelligence Constant research in optimizing the cost [29] and utility functions [30] has brought IoT algorithms into the limelight. Quest to find optimum polynomial complexity algorithms for NP and NP complete problems is one of the most prominent research studies being pursued in the domain of IoT. These NP problems are often the mathematical equivalent of various real-life problems. Deep Learning and Neural Networks [31] are the more recent endeavor toward automation of analytics. The idea behind the algorithms developed is to mimic the human neural structure at a high level. The primary component of the neural networks structures is perceptron aka artificial neuron. It is a binary classifier. Neural network is composed of various layers of perceptron with each layer having multiple perceptron, as shown in Figure 4.2. Since, mathematically, a neural network is equivalent to a graph, perceptrons in a neural network are also called nodes. There are various types of methods commonly used in Deep Learning [32] like Convolutional Neural Network used for image detection and classification, Recurrent Neural Network, Restricted Boltzmann Machine, and Autoencoders. Deep Learning models are prominently used in object classification and pattern recognition. Deep Learning has an edge over IoT because it gets better with an increase in the number of training data instances, making it a suitable choice for application in Big Data analytics. Hidden Layers Node

Node

Node

Input Layer

Output Layer Node

Node Neural Network

Figure 4.2  A typical neural network with two hidden layers.

Node

74  The Smart Cyber Ecosystem for Sustainable Development The iterative nature and hence non-linear computational complexity of various IoT and Deep Learning algorithms made it nearly impossible to carry out their implementation despite the theoretical know-how. CPU-GPU parallelism and Supercomputers gave a great impetus to the implementation of these algorithms. As algorithms and computers continue to evolve, so does the entire framework of IoT and Deep Learning.

4.2.3 Programming Languages for Artificial Intelligence There are various programming languages that are used for the implementation of AI systems. Most of them share one common feature that they are functional programming languages [33]. Some prominent examples are Lisp, Python, and, recently emerging, Julia. Python is the most used language for AI implementation these days because of the hundreds of thousands of packages and these impart versatility to Python language. The integrability with languages like C and C++ gives Python the requisite speed that it lacks due to its high-level programming syntax. In fact, the majority of the prominent IoT libraries used in Python use C for matrix and vector calculations. Python also has libraries like PyTorch and TensorFlow, which provide hosted and local runtime GPU and TPU integration. As mentioned before, GPU implementation fastens the computation. Julia, a newer programming language [34], has posed a serious threat to the prominence of Python due to its fast implementation of Linear Algebra and Calculus. Moreover, Julia has a high-level syntax too, making it easy to understand and implement. It combines strengths of C (speed) and Python (syntax) into one programming language. Since it is a relatively new language, it does not compete with the versatility of Python. Julia, however, supports easy incorporation of Python and R libraries through PyCall and RCall packages. Various other programming languages like Rust and Clojure have also emerged which aim to overcome the speed issue of Python. This indicates the constant development that is being undertaken in the development of suitable programming languages for implementation of AI, specifically for IoT and Deep Learning.

4.2.4 Outcomes of Artificial Intelligence We covered the fundamentals of IoT and how we can implement it using various programming languages and methods. The possible outcomes of usage of AI models are as follows: 1. Wider data analysis: With the usage of AI models, we can read and evaluate enormous volumes, far beyond manual capacity. Industrial scale data was millions of data instances with thousands of data features. Usage of AI fastens the process, and at the same time, larger data helps to create more stable and accurate conclusions. So, a greater analysis speed along with higher precision in conclusion makes AI highly lucrative. 2. Economic profitability: As mentioned above, firms can bypass a lot of manual work by automation and increase profit by serving a larger clientele. There might seem to be a possible loss of employment, but it is not so. Paradigm of employment changes from changes, not the employment per se. 3. Mass availability of resources and end of disparity: Earlier, the technical education that was limited to colleges and training institutions is now available

AI in Big Data and IoT  75 through mass online learning platforms at much lower cost. The mass of availability became possible only through deliverance of colossal amounts of data via the internet. The role of Big Data here is that it has helped the content creators and consumers find each other through predictive analysis. 4. Scientific research: Various research fields like bioinformatics and molecular physics require voluminous calculations and data analysis to thrive. IoT and Big Data models are being utilized in such fields. Thus, AI has become omnipresent in our lives and it is being touted as Industrial Revolution 4.0. Predictive and Behavioral Analysis Analysis of activity trends of users on public platforms like YouTube, Amazon, and Google helps these companies and the third parties with whom they share these data to formulate marketing and production strategies. Such analysis is called predictive analysis. Understanding the human behavior and its subtleties gives a greater insight into what people want, such an analysis is called behavioral analysis. To facilitate the analysis of millions (and possible billions) of users, we need AI models that crunch the Big Data to something meaningful. Usage of predictive analysis through publicly sourced data is now being used by individual content creators. They extrapolate the market trend in their field and create the content accordingly. This mutually benefits the creators, who get large audiences and consumers who get great content. Predictive analysis is a type of reinforcement learning, where the model learns with its own experiences and alters the algorithm and feature engineering accordingly. Deep Learning emerged to tackle these problems and has proved to be very promising for the same. Behavioral analysis is important toward the growth of AI as our aim is to mimic the human mind to the best possible extent.

4.3 Big Data Big Data analyses are modern science and business centric. The data is generated from online-transactions, e-mails, videos, pictures, audios, online streaming, data logs, posts, responses, data records, queries, social media, science data, data records of sensors, and mobile phone applications [35–37]. Earlier, it takes years to generate a huge volume of data but in modern time it is quite easier to create data in Exabytes (1,018 bytes) [38]. A revolutionary step is required to take the traditional technique of data analysis in Big Data by three major components which are as follows: variety, velocity, and volume [39–42]. Figure 4.3 depicts the three components of Big Data also known as three V’s of Big Data. A brief about each component of Big Data are discussed below: 1. Variety: Big Data comes from diverse sources and usually comes in three different forms: structured, semi-structured, and unstructured. A well-defined data in terms of length and format is known as Structured Data. Numbers, dates, and strings can be considered as the most common examples of structured data and are most commonly stored in databases. Structured data has

76  The Smart Cyber Ecosystem for Sustainable Development VOLUME

VARIETY

Terabyte Petabyte Exabyte Zettabyte

Unstructured Semi-Structured Structured

BIG DATA Batch New Real Time Real Time Streams VELOCITY

Figure 4.3  Three V’s of Big Data.

come to the force as a paradigm shift in technology which makes data mining easier for business insights. However, it is highly predictable, easy to organize, and can be found easily using fundamental algorithms. They contain rigid information and allow the user to enter data in some specific fields containing textual and numerical data having predefined size. In addition to it, structured data have certain protocols to access the information enclosed with the data. There are two major sources of getting structured data and they are machine-generated (sensor data, web log data, and financial data) and human generated (on click data, input data, game data) [43]. Unlike structured data, we have semi-structured data that contains semantics tags but does not reveal anything about the structure associated with relational databases. While being in the same class, they can have different types. The best example that can be quoted in this context is an email that consists of the collection of XML and other markup languages [44]. On the other hand, we have unstructured data that is not organized in any form (size and format) and has no associated model with it. Unstructured data is stored in data lakes and includes a huge volume of texts and multiple types of data. Books, health records, satellite images, and adobe pdf files are the best examples that can be inferred to understand unstructured data [45]. 2. Volume: Volume or the size of data plays a significant role in Big Data as its name suggests that the size of the data should be huge to get referred as Big Data. The size/volume can take up to any size from terabytes (1012 bytes) to exabytes (1021 bytes) [46]. 3. Velocity: One of the most incumbent traits for the modern day-to-day technology as it is responsible for time limited flow of data and processes. Most of the applications such as Facebook and YouTube require a high speed of data chunks to be available for a continuous streaming of videos and even in voice over internet protocol we require a lot of continuous data flow to make an audio/video call over the internet [47].

AI in Big Data and IoT  77

4.3.1 Artificial Intelligence Methods for Big Data Both AI and Big Data are prominent and utilitarian to technology in the present era. AI is an existing form more than a decade now, but Big Data came into a lame light a couple of years ago. Earlier, computers were having storage capability but the strength to inspect this data is provided by Big Data. We can say that together Big Data and AI are technologies that empower ML by continuously updating the data records, with human intervention and recursive experiments for the same [48]. Several organizations consider that organizational data will be revolutionized by AI tools. ML helps the system to itself predict the outcome and is an advanced version of AI. Big Data examines organizations existing data and furnishes meaningful insights from the same data set. Now, the question here arises how Big Data helps in AI experiments? As it is very clear that AI will make a human intervention with machines almost negligible, considering AI has all functionalities of ML that will create a technology to take over human jobs known as robotics these days. The human interference will be reduced due to AI expansion and this though is incomplete without the involvement of Big Data. Machines can take decisions on behalf of the data provided to them but cannot make decisions based on emotions, which later was overcome by Data Scientists. This technical development helped a lot in the pharmaceutical sector where machines can tell the name of medicines based on the emotions of the patient. So, the unification of AI and Big Data can not only examine the requirements that the users buy and inhibit some constraints and protocols of that sector [49]. With every passing day, a new cost-efficient technology is introduced in the market. As a result, the demand for that product will be high in the market. Even in a country like India with its varied culture, language, and religion the products will be acquired with the similar potential. With that, the provider will have to bring the comparable solutions going with the customer needs. Big Data technology will help the organizations by providing appropriate solutions to their problems with respect to language and religion; on the other hand, ML will provide them solutions based on emotional values attached [50]. At present, the market does not know the requirements of the customers but with time this problem will also be solved through the help of AI and Big Data technology.

4.3.2 Industry Perspective of Big Data Since we all know when data is present in large amounts, then it is referred to as Big Data and Big Data is very useful in understanding trends in the data and to maintain records for better yield. In earlier times, it was difficult to analyze the shifts which can be resourceful to the company. This problem arose because of a lack of data consistency and the manual approach of keeping records, in which the data can be easily manipulated and uploaded in the record. The technology swept too fast and the requirement of large data was increased within a quick span. The data processing speed was also slow in earlier systems [51]. Moreover, security and data privacy, data access and information sharing, processing and storage issues, analytical challenges, and technical challenges were few more challenges that were faced by people during the beginning of the Big Data era [52]. These days, we can see several applications where Big Data plays an incumbent role in proper functioning and efficient usage.

78  The Smart Cyber Ecosystem for Sustainable Development

EDUCATION ASTRONOMY

ANALYTICS

MEDICAL

BIG DATA

INDUSTRY

WEATHER

Figure 4.4  Applications of Big Data.

Figure 4.4 shows some of the applications of Big Data in the present era. The image itself depicts that the applications of Big Data are now not only restricted to some fields but have diversified into many other fields.

4.3.2.1 In Medical Field In the healthcare department, a lot of times we do require data that we can store at a single place to get insight knowledge of things. With the help of Big Data, things become much smoother in the medical field to work upon because of its variety, veracity, and volume. Big Data is becoming a most favorable option for the researchers to help in reducing the cost of care measurement, providing excellent clinical support and to manage the population of the patients at high risk. Big Data is helping a lot in diagnosing cancer and risk of cardiac arrest at early stages. It also contributes to monitoring the heart rate, blood pressure, and respiratory rate; any change in the record can alert the doctors to take necessary actions upon that patient [53].

4.3.2.2 In Meteorological Department One might wonder how Big Data can help the meteorological department to analyze things. Well, the answer is yes, it helps in many ways to help the department to analyze the currents and breeze and to sense or monitor the daily atmospheric conditions through various sensors which help to record accurate data. The weather has been an issue of worry for

AI in Big Data and IoT  79 everyone in the past some years which was changing day by day, and it was very necessary to understand the shifts and find its applications in weather forecasting. Big Data provides a great volume of data for this purpose and helps the system to forecast the weather conditions and hence help many farmers to check the yield of crops [54, 55].

4.3.2.3 In Industrial/Corporate Applications and Analytics All industrial and corporate fields require a huge volume of data to assess the growth of their company to predict the annual revenue from the applications and to judge the production and consumption of the raw material and products to mold its yield accordingly. Big Data analysis studies the patterns in the variations and predicts the production/ consumption of units in the year. A Big Data framework can be utilized to assess electric power data which integrates the real- time data and the previous data. This provides us a framework to improve the performance and efficiency of the system [56]. Similarly, many other companies are also paying much attention to using the alluring technology of Big Data. The government runs various federation programs to collect data from various levels and store it in a single place [57]. For example, Aadhar card, for a very big population, it is quite difficult to keep a record of 1.3 crore population of a huge country like India so it will be better if we use Big Data analysis to keep our records up to date. Similarly, in the USA, the government keeps an eye on every US national through the Social Security Number (SSN) to know the status of the person at any time anywhere and will reach the needy in case of emergency.

4.3.2.4 In Education Big Data in the education sector will help in improving the results of the students; dropout rates at schools and colleges will also be reduced. Schools/universities can use predictive analysis over the data which is being provided to get an overall insight into the student’s upcoming performance. Big Data in the education sector brings unprecedented opportunities to reach out and instruct students in new techniques (customized programs) which will help students in understanding concepts more thoroughly and memorize them easily. Overall, as shown in Figure 4.5, Big Data helps the education sector in four major steps: improve student result, customize programs, result in dropouts, and targeted international recruiting better learning will help most of the students to score well in exams and results in achieving good grades and hence reduces the dropout chances. Big Data also produces an influx in the data while processing the applications online, and hence, this situation is usually targeted by foreign universities in which many students apply throughout the globe. Big Data has the potential to revolutionize the learning industry in the coming years. Smarter students will have a positive impact on organizations and society. Therefore, it is time we embrace Big Data in the education sector [58].

4.3.2.5 In Astronomy Applications of Big Data can be seen in space and satellite communication systems in which we use robots wherever the accessibility of humans is difficult. So, we design robots which

80  The Smart Cyber Ecosystem for Sustainable Development

IMPROVE STUDENT RESULT

REDUCE DROPOUTS

CUSTOMISE PROGRAMS

TARGETED INTERNATIONAL RECRUITING

Figure 4.5  Major contribution of Big Data in education.

are well equipped with the latest technology so that they can share the information about celestial bodies from time to time and help mankind to learn new things more easily. The data can be of any type of pictures, text, or numbers [59]. Although image and signal processing is in practice, we hope that, in future, we can merge Big Data analytics with AI to get some better results and to study the variations in other planets too.

4.4 Internet of Things IoT is another emerging technology that is used to control everything from anywhere anytime. IoT interconnects several sensors together to form a resourceful device that can be used for the welfare of society. Rather than people-to-people communication, IoT gives more emphasis on the machine to machine communication [60]. It plays an essential role in the field of robotics and can be merged with ML for better results. On the unification of IoT with AI it results in several useful applications that reduce the workload of humans to perform the functions to complete the work. For example, mostly every website is using a chatbot under customer care tag so that the basic queries and general information can be given to the end-user by the system itself if he further needs any assistance then he can call the given customer care number to clarify his/her query. Sensors like IR sensors, temperature sensors, optical sensors, proximity sensors, etc., provide a good volume of data to the IoT-based system which can be further optimized by using Big Data practices to merge the present technology with ML and on analysis it predicts the output, such type of systems are known as Analytics of Things (AoT). AoT can be defined as “The analysis of data generated by the IoT is referred to as Analytics of Things (AoT)” [61].

AI in Big Data and IoT  81

4.4.1 Interconnection of IoT With AoT The IoT has provided us a platform to work independently and provided us with flexibility to do work from anywhere provided the machine has an internet connection. Moreover, it is a very easy technique that anyone without even having prior experience can utilize this technology. We have seen several applications such as Magic Mirror which is itself an IoTbased project and its function is to show all the necessary information which can be seen on the screen of a mobile phone and one can also have personalized messages on the screen; it can show news, real-time weather, time, and other several important information on the screen of the mirror. The future scope of this application is we can connect a microphone to it along with speakers and it can behave as a competitor to Google Home or Amazon Alexa. Here, the system will also hear several voices which the machine needs to remember while performing the commands. The complexity of data has again risen, and we need to apply Big Data analytics on the data provided by the sensors so that the time is reduced and hence the process gets smooth [62]. There are four types of data analytics that can be helpful with IoT to earn profit and they are as follows: 1. Streaming Analytics: This set of analysis is also referred to as event stream processing and analyzes a huge volume of in-motion data. The best example which can be quoted is the systems which need real-time data like weather monitoring and air pollution tracker. 2. Spatial Analytics: This type of analytics is highly used for studying the geographic locations to observe physical patterns in between the objects. Location-based IoT products (smart parking application) can be used as examples for this analytic method. 3. Time Series Analytics: As the name indicates, this form of data analytics is based upon time which is analyzed to reveal various trends and shifts in patterns. Applications like health monitoring systems can be inferred for studying such methods. 4. Perspective Analysis: This form of analytics is combined with the descriptive and predictive analysis. It is applied to understand the best actions which can be taken in accordance with a particular situation. Commercial IoT applications can be referred as bestexample [63].

4.4.2 Difference Between IIoT and IoT No doubt the IoT has proved itself in every field of development and making the world which we know today, and Industrial IoT (IIoT) is derived from the parent, IoT. If we see the comparison between both, they are quite similar in some respects and, at the same time, quite different in many aspects. IoT is the interconnection of several components which results in a product whereas the IIoT is also Interconnection of several components and machines which results in a useful technology and can be utilized by numerous industries. IoT applications are mainly designed for an individual which can be used at homes and offices for personal usage, whereas the IIoT is used for commercial purposes only. Security has never been a major concern in IoT as it is based on individual needs, but we cannot

82  The Smart Cyber Ecosystem for Sustainable Development take such risk in IIoT where the crucial information of the company is at high risk. IoT uses applications with low risk impacts; on the other hand, IIoT utilizes more sensitive and precise components which makes IIoT products expensive as compared to IoT products [64].

4.4.3 Industrial Approach for IoT Unification of two technologies, IoT and Big Data, has resulted in many useful applications which are contributing in one or the other way for the social welfare of mankind. Various applications of the IIoT are mostly concerned with networks. Here, we are not only talking about the physical networks in industries but also includes the representation of products, processes, and services such as 3D modeling or physical behavioral models. In 2015, IoT was declared as the most publicized technology [65]. IIoT has expanded from the integration of numerous technologies. There are some examples of IIoT which some of the companies are using: 1. MAN: A company which deals with the manufacturing of trucks and buses. It tends to provide support with the tracker to track the faults in the engine and reduces the labor of the mechanic to fixit [66]. 2. Siemens: A well-established German-based multinational company (MNC) which produces automated products for grand brands such as British Motor Works (BMW). Siemens made us familiar with an Operating System (OS) called MindSphere, which basically collects the data from various different crucial departments and processes them with rich analytics to furnish one of a unique product [67]. 3. Caterpillar (CAT): CAT is an American machinery and equipment firm, which operate machines by augmented reality (AR) applications. With intelligent sensors and network capabilities, it optimizes and monitor closely. Solutions provided by CAT are forefront working for the customer’s satisfaction [68]. 4. Tesla: Another American firm which is specialized in the manufacturing of electric vehicles and its energy restoration products [69].

4.5 Technical Shift in AI, Big Data, and IoT Altogether, we can say AI, Big Data, and IoT have profound impact in shaping the world in which we are living now and helps in fulfilling the desire of modern needs. The next generation of work, between advances and interconnectivity in AI and IoT, depicts that the world is changing and reduces the man labor in various domains. IoT is supporting the changes in society which we are already familiar with. Products are getting converted into services, instead of purchasing cars people are looking for carpooling. Airlines are looking forward to paying hourly wages while the engine is propelling. If the service requires a physical asset, then it is considered to be an ideal case where the asset can work independently, which is resolved by IoT that allows users to monitor and collect data and various other parameters which can affect the automation of the asset without even visiting.

AI in Big Data and IoT  83

4.5.1 Industries Shifting to AI-Enabled Big Data Analytics These days, businesses revolve around data and advanced technologies like AI and Big Data that are developing like an essential player in the changing of modern businesses. Businesses in the present scenario utilize these technologies to make decisions, to analyze patterns, and to make predictions about the revenue and demand/supply of products. On applying analytics in Big Data, businesses can get some productive knowledge. Whereas AI is being used to get insight knowledge about the customers behavior to understand the proceeding shifts which can make the firm better than its competitors [70, 71]. The benefits of AI-enabled Big Data analytics is described below: 1. Inventory Management (IM): IM is associated with the collection of actions engaged to complete a task. Actions like push or pop fields, update fields, stock maintenance, and monitoring all the databases. Through AI technology, businesses can apply prophetic analysis on the inventory data, which will help in extracting useful information from records and tells us about the current and future demand of the product; moreover, it makes our inventory smart and gives knowledge about the behavior of customers. The Business owners can understand the real-time demand and future demand of the product and manufacture it accordingly. Big Data will help organizations to predict the upcoming trends and about the demand/sales forecasting [72]. 2. Employee Engagement: Nowadays, many people are not working in their field of area due to lack of opportunities and they need to work there by going against their will. They are not even properly associated with the work and they are not even satisfied with their jobs. It is every business owner’s responsibility to have proper knowledge about the obstacles one can face in the path of success and the strategies which could be applied to overcome all these barriers. Recently, AI has emerged as a powerful tool to help in business engagements. It provides some tools like Tango work which helps employees to discuss everything which they can discuss with their boss. Thus, it provides a friendly ambience in and around the office [73]. 3. Customers Behavior: Beyond belief, predictive analysis of AI helps a firm to understand numerous requirements of the customers. AI-enabled applications such as GPU database keep customer information and predict their behavior according to it. Client preferences are given priority to enhance their experience. AI in marketing helps businesses to target appropriate audiences for their advertising. AI will help in targeting the best customers through mails and messages whereas Big Data will be used internally to monitor the employee’s performance and designing accurate delivery models [74, 75]. Few organizations are using this technology to develop smart workspace. These workspaces have a long chain of interconnected sensors to monitor their staff. These sensors can track the voice, stress, and many other factors [76]. 4. Fraud Prevention and Smart Recruitment: In the era of digitalization and in the path of becoming a developed country, digitalization is very important, but with this upcoming trend, some challenges are also associated among

84  The Smart Cyber Ecosystem for Sustainable Development which fraud prevention is the incumbent. AI-powered algorithms identify the fraudulent behaviors based on the Big Data based on analyzed historical data [77]. Another challenging fact about digitalization is the rise of competitors in the market. AI algorithms have made an easier platform for HR’s to recruit people. It has reduced the effort and time of the selection process now they can perform these tasks just in a few seconds [78]. Thus, we can see how AI and Big Data are transforming the future of businesses with incredible tools and techniques. The upcoming business future is bright and clear with automation and real-time operations.

4.5.2 Industries Shifting to AI-Powered IoT Devices IoT devices have the capability of generating a large volume of data with the help of sensors that can be used by AI, and all we need to do is to construct an IoT model or algorithm by consuming that data. Since the IoT platform provides an interface to collect information and data from various sources (sensors, devices, etc.), they can easily deploy to AI systems. The importance of AI here is to make quick decisions by studying and analyzing the shifts and trends in the data set provided to it by IoT. There are a few advantages of unifying AI with IoT which is briefly described below [79]: 1. Avoiding unplanned downfall: Using predictive analysis the firm cannot face a plunge in the selling of the product. If it is updated regularly, then the owner will never be in loss. 2. Increasing operational efficiency: AI-enabled systems will help the user to predict the operating conditions and identify the parameters to be adjusted for a long-term investment. 3. Enabling improved products and services: Unification of AI and IoT is helping in creating many new featured services which were never imagined and used before. It has the utmost contribution in making this world automatic from vehicles to home. Everything is getting automatic now with the help of AI-powered IoT devices. 4. Enhancing risk management: This unique and beneficial combination of Technology is helping various organizations for understanding and predicting a variety of risks and by providing a solution for managing them in a better way. It enables better opportunities for companies in terms of cutting edge technology. Azure IoT, AWS, and Google Cloud IoT are some examples of it [80].

4.5.3 Statistical Data of These Shifts The following statistical information available in the Figure 4.6, represented as a bar graph throws light upon the total number of active device connections available worldwide with IoT connectivity in 5 years before 2020 and shows a predictive analysis about the upcoming 5 years. According to the provided information, it is noticeable that in 2015, there were a total of 13.9 billion bases in which only 3.8 billion bases were having IoT connectivity because

Number of Global Active connections (installed base) in Bn

AI in Big Data and IoT  85 Total Number of Active Device Connections WorldWide 40 35 30 25 20 15 10 5 0

2015

2016

2017

2018 IoT

2019

2020

Non-IoT

2021

2022

2023

2024

2025

Average growth

Figure 4.6  Statistical data of AI-powered IoT [81].

the IoT technology was just started to begin this year. A gradual increase in the users of IoT devices can be seen in the years 2016 and 2017 can be seen in the above figure. By 2018, approximately, 7.0 billion people started utilizing the advanced technology provided by IoT to its users out of 17.8 billion users. In the year 2020, there are 21.2 billion users who are using connections and roughly about 10 billion users are using IoT-based devices for their home and offices. By observing the statistical data of last 5 years, it can be predicted that there will not be any downfall in the users of IoT technology in upcoming 5 years, and we will see that by 2025, the total users will be soared to 34.2 billion out of which 21.5 billion people will be using the trending technology of IoT-powered AI devices, which is more than 50% of the total active connections. To recapitulate, we can say that at the beginning of the decade, i.e., 2015, only 30% (approximately) of the total connections were using IoT-powered devices, but later on, it can be seen that the IoT-powered devices have brought a revolution with time and the consumption of these devices have inclined gradually to roughly around 75% which indicates that we are developing the technology along with our lifestyle with every passing year.

4.6 Conclusion In this chapter, we are focusing on the technological shift caused by the three astonishing technologies which are AI, Big Data, and the IoT. All these technologies are equally important in shaping the modern world. On one side, AI is giving the intelligence to the machines and allowing the system to itself learn and predict the upcoming shifts based on past trends. AI alone is itself a prominent technology, but when it is used with Big Data and IoT, the efficiency of the system is increased than before. When AI is empowered with Big Data, then more optimized and better results can be obtained from it, which are very helpful for reputed companies to understand the behavior of the market, whereas both the IoT and AI

86  The Smart Cyber Ecosystem for Sustainable Development technologies are robust and are efficient in making business smarter. Integrating these two technologies together, it will enable organizations to gain even higher transformation. There are numerous areas where the applications of this unification can be utilized. Of course, this unification is expensive but worth its cost. This integration is not only helpful in business only, but its applications can be seen in security and access devices, emotional analysis, and facial recognition. AI has brought the fiction world into reality, and its several applications which we can use at present in the automation industry which made possible to run driverless cars and trains, in detecting the frauds even fake news too, in healthcare systems to AI along with IoT and robotics are contributing a lot considering the fact that AI can do surgeries with better precision than humans and, even in situations like COVID-19, robots are being used for taking care of infected patients, drone are being used to monitor the situation of lockdown, and found to be the best companion to humans as, in some part of the world, drones are found to take pets on a regular walk. In various missions where humans cannot reach, this combination of technology is very resourceful to us in such applications.

References 1. Kristyn, A., Kharlamov, M., Prokhorov, S., Shcherbinin, D., Application of Artificial Intelligence Systems in the Process of Crew Training. 2018 International Conference on Artificial Intelligence Applications and Innovations (IC-AIAI), Nicosia, Cyprus, pp. 55–59, 2018. 2. Ahmad, A.S. and Sumari, A.D.W., Cognitive artificial intelligence: Brain-inspired intelligent computation in artificial intelligence. 2017 Computing Conference, London, pp. 135–141, 2017. 3. Arsenijevic, U. and Jovic, M., Artificial Intelligence Marketing: Chatbots. 2019 International Conference on Artificial Intelligence: Applications and Innovations (IC-AIAI), Belgrade, Serbia, pp. 19–193, 2019. 4. Xian, L., Artificial intelligence and modern sports education technology. 2010 International Conference on Artificial Intelligence and Education (ICAIE), Hangzhou, pp. 772–776, 2010. 5. Liu, Y. and Tang, P., The Prospect for the Application of the Surgical Navigation System Based on Artificial Intelligence and Augmented Reality. 2018 IEEE International Conference on Artificial Intelligence and Virtual Reality (AIVR), Taichung, Taiwan, pp. 244–246, 2018. 6. Gonzalez-Alonso, P., Vilar, R., Lupiañez-Villanueva, F., Meeting Technology and Methodology into Health Big Data Analytics Scenarios. 2017 IEEE 30th International Symposium on Computer-Based Medical Systems (CBMS), Thessaloniki, pp. 284–285, 2017. 7. Panda, M., Ali, S.M., Panda, S.K., Big data in healthcare: A mobile based solution. 2017 International Conference on Big Data Analytics and Computational Intelligence (ICBDAC), Chirala, pp. 149–152, 2017. 8. Philip Chen, C.L. and Zhang, C.-Y., Data-intensive applications, challenges, techniques, and technologies: A survey on Big Data. Information Sciences. 275, 314–347, 2014, https://doi. org/10.1016/j.ins.2014.01.015. 9. Yin, J. and Zhao, D., Data confidentiality challenges in big data applications. 2015 IEEE International Conference on Big Data (Big Data), Santa Clara, CA, pp. 2886–2888, 2015. 10. Roy Chowdhury, A., IoT and Robotics: a synergy. PeerJ Preprints, 5, e2760v1, 2017. https://doi. org/10.7287/peerj.preprints.2760v1. 11. Talwana, J.C. and Hua, H.J., Smart World of Internet of Things (IoT) and Its Security Concerns. 2016 IEEE International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data (SmartData), Chengdu, pp. 240–245, 2016.

AI in Big Data and IoT  87 12. Biswas, A.R. and Giaffreda, R., IoT and cloud convergence: Opportunities and challenges. 2014 IEEE World Forum on Internet of Things (WF-IoT), Seoul, pp. 375–376, 2014. 13. Nickolls, J. and J. Dally, W., The GPU Computing Era. IEEE Micro, 30, 2, 56–69, March-April 2010. 14. O’Leary, D.E., Artificial Intelligence and Big Data. IEEE Intell. Syst., 28, 2, 96–99, March-April 2013. 15. Franco-Arcega, A. et al., Decision tree-based Classifiers for large Datasets. Computación y Sistemas, 17, 1, 95–102, 2013. 16. Granger, C.W.J., Prediction with a Generalized Cost of Error Function. J. Oper. Res. Soc., 20, 2, 199–207, 1969. 17. Levy, H. and Markowitz, H.M., Approximating Expected Utility by a Function of Mean and Variance. Am. Econ. Rev., 69, 3, 308–317, 1979. JSTOR, www.jstor.org/stable/1807366. 18. Cruz, J.A. and Wishart, D.S., Applications of Machine Learning in Cancer Prediction and Prognosis. Cancer Inf., 2, 59–76, Jan. 2006. 19. Shabtai, A., Fledel, Y., Elovici, Y., Automated Static Code Analysis for Classifying Android Applications Using Machine Learning. 2010 International Conference on Computational Intelligence and Security, Nanning, pp. 329–333, 2010. 20. Gabrielsson, P. and Johansson, U., High-Frequency Equity Index Futures Trading Using Recurrent Reinforcement Learning with Candlesticks. 2015 IEEE Symposium Series on Computational Intelligence, Cape Town, pp. 734–741, 2015. 21. Pandey, S. and Tokekar, V., Prominence of MapReduce in Big Data Processing, 2014 Fourth International Conference on Communication Systems and Network Technologies, IEEE, pp. 555–560, 2014. 22. Tan, W., Gan, F., Chang, T., Using normal quantile plot to select an appropriate transformation to achieve normality. Comput. Stat. Data Anal., 45, 3, 609–619, 2004. 23. Qu, L., Li, L., Zhang, Y., Hu, J., PPCA-Based Missing Data Imputation for Traffic Flow Volume: A Systematical Approach. IEEE Trans. Intell. Transp. Syst., 10, 3, 512–522, Sept. 2009. 24. Hadsell, R., Chopra, S., LeCun, Y., Dimensionality Reduction by Learning an Invariant Mapping. 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’06), New York, NY, USA, pp. 1735–1742, 2006. 25. Lin, Y., Liu, T., Fuh, C., Multiple Kernel Learning for Dimensionality Reduction. IEEE Trans. Pattern Anal. Mach. Intell., 33, 6, 1147–1160, June 2011. 26. Harsanyi, J.C. and Chang, C. -., Hyperspectral image classification and dimensionality reduction: an orthogonal subspace projection approach. IEEE Trans. Geosci. Remote Sens., 32, 4, 779–785, July 1994. 27. Stulp, F. and Sigaud, O., Many regression algorithms, one unified model:A review. Neural Networks, 69, 60–79, 2015. 28. Kotsiantis, S.B., Zaharakis, I.D., Pintelas, P.E., Machine learning: a review of classification and combining techniques. Artif. Intell. Rev., 26, 159–190, 2006. https://doi.org/10.1007/ s10462-007-9052-3. 29. Bengio, Y. and LeCun, Y., Scaling learning algorithms towards AI, in: Large-scale kernel machines, vol. 34, no. 5, pp. 1–41, 2007. 30. Hibbard, B., Model-based Utility Functions. J. Artif. Gen. Intell., 3, 1, 1–24, 2012. 31. LeCun, Y., Bengio, Y., Hinton, G., Deep learning. Nature, 521, 436–444, 2015. https://doi. org/10.1038/nature14539. 32. Voulodimos, A., Doulamis, N., Doulamis, A., Protopapadakis, E., Deep Learning for Computer Vision: A Brief Review. Comput. Intell. Neurosci., 2018, 1–13, 2018. 33. Priestley, M., AI, and the Origins of the Functional Programming Language Style. Minds Mach., 27, 449–472, 2017. https://doi.org/10.1007/s11023-017-9432-7.

88  The Smart Cyber Ecosystem for Sustainable Development 34. Bezanson, J., Edelman, A., Karpinski, S., Shah, V.B., Julia: A Fresh Approach to Numerical Computing. SIAM Rev., 59, 1, 65–98, 2017. 35. Sagiroglu, S. and Sinanc, D., Big data: A review. 2013 International Conference on Collaboration Technologies and Systems (CTS), San Diego, CA, pp. 42–47, 2013. 36. Eaton, C., Deroos, D., Deutsch, T., Lapis, G., Zikopoulos, P.C., Understanding Big Data: Analytics for Enterprise Class Hadoop and Streaming Data, McGraw-Hill Companies, 978–0-07-179053-6, United States of America, 2012. 37. Schneider, R.D., Hadoop for Dummies Special Edition, John Wiley & Sons, Canada,  978–1118-25051-8,2012. 38. Intel IT Center, Planning Guide: Getting Started with Hadoop, in: Steps IT Managers Can Take to Move Forward with Big Data Analytics, June 2012, http://www.intel.com/content/dam/www/ public/us/en/documents/guides/getting-started-with- hadoop-planning-guide.pdf. 39. Gerhardt, B., Griffin, K., Klemann, R., Unlocking Value in the Fragmented World of Big Data Analytics, in: CiscoInternet Business Solutions Group, June 2012, http://www.cisco.com/web/ about/ac79/docs/sp/InformationInfomediaries.pdf. 40. Singh, S. and Singh, N., Big Data Analytics. 2012 International Conference on Communication, Information & Computing Technology Mumbai India, IEEE, October2011. 41. Sandryhaila, A. and Moura, J.M.F., Big Data Analysis with Signal Processing on Graphs: Representation and processing of massive data sets with irregular structure. IEEE Signal Process. Mag., 31, 5, 80–90, Sept. 2014. 42. Ram, J., Zhang, C., Koronios, A., The Implications of Big Data Analytics on Business Intelligence: A Qualitative Study in China. Procedia Comput. Sci., 87, 221–226, 2016. https:// doi.org/10.1016/j.procs.2016.05.152. 43. Subramaniyaswamy, V., Vijayakumar, V., Logesh, R., Indragandhi, V., Unstructured Data Analysis on Big Data Using Map Reduce. Procedia Comput. Sci., 50, 456–465, 2015. https://doi. org/10.1016/j.procs.2015.04.015. 44. Madden, S., From Databases to Big Data. IEEE Internet Comput., 16, 4–6, June 2012. 45. Chardonnens, T., Cudre-Mauroux, P., Grund, M., Perroud, B., Big data analytics on high Velocity streams: A case study. 2013 IEEE International Conference on Big Data, Silicon Valley, CA, pp. 784–787, 2013. 46. https://www.datamation.com/big-data/big-data-vs.-artificial-intelligence.html 47. Yin, J. and Zhao, D., Data confidentiality challenges in big data applications. 2015 IEEE International Conference on Big Data (Big Data), Santa Clara, CA, pp. 2886–2888, 2015. 48. Hoehndorf, R. and Queralt-Rosinach, N., Data Science and Symbolic AI: Synergies, Challenges and Opportunities. Data Sci., 27 –38, 1, 1–2, Jan. 2017. 49. Zhang, H.L. and Lau, H.C., Agent-based Problem Solving Methods in Big Data Environment. Web Intelligence and Agent Systems, 12, 4, 343–345, 1 Jan. 2014. 50. Hussain, M. and Manhas, Dr., Artificial Intelligence for Big Data: Potential and Relevance. Int. Acad. Eng. Med. Res., 1–5, 2016. 51. https://www.piesync.com/blog/top-5-problems-with-big-data-and-how-to-solve-them/ 52. Katal, A., Wazid, M., Goudar, R.H., Big data: Issues, challenges, tools and Good practices. 2013 Sixth International Conference on Contemporary Computing (IC3), Noida, pp. 404–409, 2013. 53. Rahman, F., Slepian, M., Mitra, A., A novel big-data processing framwork for healthcare applications: Big-data-healthcare-in-a-box. 2016 IEEE International Conference on Big Data (Big Data), Washington, DC, pp. 3548–3555, 2016. 54. Jain, H. and Jain, R., Big data in weather forecasting: Applications and challenges. 2017 International Conference on Big Data Analytics and Computational Intelligence (ICBDAC), Chirala, pp. 138–142, 2017.

AI in Big Data and IoT  89 55. Juneja, A. and Das, N.N., Big Data Quality Framework: Pre-Processing Data in Weather Monitoring Application. 2019 International Conference on Machine Learning, Big Data, Cloud and Parallel Computing (COMITCon), Faridabad, India, pp. 559–563, 2019. 56. Liu, H., Huang, F., Li, H., Liu, W., Wang, T., A Big Data Framework for Electric Power Data Quality Assessment. 2017 14th Web Information Systems and Applications Conference (WISA), Liuzhou, pp. 289–292, 2017. 57. Aron, J.L. and Niemann, B., Sharing best practices for the implementation of Big Data applications in government and science communities. 2014 IEEE International Conference on Big Data (Big Data), Washington, DC, pp. 8–10, 2014. 58. Hadwer, A.A., Gillis, D., Rezania, D., Big Data Analytics for Higher Education in The Cloud Era. 2019 IEEE 4th International Conference on Big Data Analytics (ICBDA), Suzhou, China, pp. 203–207, 2019. 59. Shao, J., Xu, D., Feng, C., Chi, M., Big data challenges in China centre for resources satellite data and application. 2015 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), Tokyo, pp. 1–4, 2015. 60. Yadav, E.P., Mittal, E.A., Yadav, H., IoT: Challenges and Issues in Indian Perspective. 2018 3rd International Conference On Internet of Things: Smart Innovation and Usages (IoT- SIU), Bhimtal, pp. 1–5, 2018. 61. Bonomi, F., Milito, R., Natarajan, P., Zhu, J., Fog Computing: A Platform for Internet of Things and Analytics, in: Big Data and Internet of Things: A Roadmap for Smart Environments. Studies in Computational Intelligence, vol. 546, N. Bessis and C. Dobre (Eds.), Springer, Cham, 2014. 62. Yadav, P. and Vishwakarma, S., Application of Internet of Things and Big Data towards a Smart City. 2018 3rd International Conference On Internet of Things: Smart Innovation and Usages (IoT-SIU), Bhimtal, pp. 1–5, 2018. 63. Ahmed, E., Yaqoob, I., Targio Hashem, I.A., Khan, I., Abdalla Ahmed, A., II, Imran, M., Vasilakos, A.V., The role of big data analytics in Internet of Things. Comput. Networks, 129, 2, 459–471, 2017. https://doi.org/10.1016/j.comnet.2017.06.013. 64. https://www.hitechnectar.com/blogs/examples-industrial-internet-of-things/. Accessed on 25 May 2020 65. Gartner, Gartner’s 2015 hype cycle for emerging technologies identifies the computing innovations that organizations should monitor, Article from Gartner's Newsroom, 2015. http://www. gartner.com/newsroom/id/3114217. Accessed 25 May 2020. 66. Official website of MAN https://www.mantruckandbus.com/en/index.html. Accessed on 26 May 2020. 67. Official website of Siemens. https://new.siemens.com/in/en/company/topic-areas/intelligent-​ infrastructure/buildings.html. Accessed 26 May 2020. 68. The official website of Caterpillar (CAT). https://www.caterpillar.com/en/company.html. Accessed on 26 May 2020. 69. Official website of Tesla under about section.https://www.tesla.com/about. Accessed on 26 May 2020. 70. Lee, J., Davari, H., Singh, J., Vibhor Pandhare,Industrial Artificial Intelligence for industry 4.0-based manufacturing systems. Manuf. Lett., 18, 20– 23, 2018. https://doi.org/10.1016/j. mfglet.2018.09.002. 71. Zhong, R.Y., Xu, X., Klotz, E., Newman, S.T., Intelligent Manufacturing in the Context of Industry 4.0: A Review. Engineering, 3, 5, 616–630, 2017. https://doi.org/10.1016/J.ENG.2017.05.015. 72. Zeng, D. and Lusch, R., Big Data Analytics: Perspective Shifting from Transactions to Ecosystems. IEEE Intell. Syst., 28, 2, 2–5, March-April 2013. 73. Ravi, V. and Kamaruddin, S., Big Data Analytics Enabled Smart Financial Services: Opportunities and Challenges, in: Big Data Analytics. BDA 2017. Lecture Notes in Computer

90  The Smart Cyber Ecosystem for Sustainable Development Science, vol. 10721, P. Reddy, A. Sureka, S. Chakravarthy, S. Bhalla (Eds.), Springer, Cham, 2017, https://doi.org/10.1007/978-3-319-72413-3_2. 74. Kim, J.K., Song, H.S., Kim, T.S., Kim, H.K., Detecting the change of customer behavior based on decision tree analysis. Expert Syst., 22, 193–205, 2005. 75. Hernández, B., Jiménez, J., José Martín, M., Customer behavior in electronic commerce: The moderating effect of e-purchasing experience. J. Bus. Res., 63, 9–10, 964–971, 2010. https://doi. org/10.1016/j.jbusres.2009.01.019. 76. Yao, X., Zhou, J., Zhang, J., Boër, C.R., From Intelligent Manufacturing to Smart Manufacturing for Industry 4.0 Driven by Next Generation Artificial Intelligenceand Further On. 2017 5th International Conference on Enterprise Systems (ES), Beijing, pp. 311–318, 2017. 77. Vaithianathasamy, S., AI vs AI: fraudsters turn defensive technology into an attack tool. Computer Fraud & Security, 8, 6–8, 2019. https://doi.org/10.1016/S1361-3723(19)30083-1. 78. Duan, Y., Edwards, J.S., Dwivedi, Y.K., Artificial intelligence for decision making in the era of Big Data – evolution, challenges and research agenda. Int. J. Inf. Manage., 48, 63–71, 2019. https://doi.org/10.1016/j.ijinfomgt.2019.01.021. 79. Ma, Y., Ping, K., Wu, C., Chen, L., Shi, H., Chong, D., Artificial Intelligence powered Internet of Things and smart public service, in: Library Hi Tech, vol. 38, pp. 165– 179, 2019, https://doi. org/10.1108/LHT-12-2017-0274. 80. https://www.wired.com/brandlab/2018/05/bringing-power-ai-internet-things/.Accessed on 28 May 2020. 81. https://iot-analytics.com/guide-iot-innovation-download-form/. Accessed on 28 May 2020.

5 IoT’s Data Processing Using Spark Ankita Bansal* and Aditya Atri Netaji Subhas University of Technology, Delhi, India 2 Netaji Subhas Institute of Technology, Delhi, India

1

Abstract

Large volume of structured and unstructured data known as Big Data requires efficient frameworks and software techniques for processing because they cannot be processed using traditional database methods. One well-known system for Big Data processing is Spark. MapReduce technology of the Hadoop was used for batch processing embedded in cluster computing. In order to help Hadoop work faster, the Spark was introduced. Spark has its own processing engine which uses distributed file storage of Hadoop and cloud storage of data. Spark’s API conforms to the type of data and its associated processing required. Spark also provides functionalities and tools for processing of queries, graphs, and machine learning algorithms. Spark SQL is very important and used for processing of queries included in the framework of Spark and hence maintaining the storage of large datasets on the cloud. Spark also performs operations on the input data taken from various different data sources. In order to maintain and create data frames, in-built functions are used by Spark. Keywords:  RDD, DataFrames, datasets, spark SQL, SQLContext, hive tables, JSON, parquet files, data sources, hadoop, MapReduce, cloud, big data, spark, cluster computing, spark API

5.1 Introduction In this chapter, we will start with the basics of three sets of APIs in Apache Spark 2.0— RDDs, DataFrames, and Datasets, which are fabricated to inculcate big data processing even easier for a wider set of audience [1]. Majorly, the focus will be on DataFrames and Datasets, as they are integrated together to ensure learning diverse concepts and grasp techniques to approach structured data and do cloud computations [2]. Because Spark provides APIs as domain-specific language constructs for higher-level abstraction. Spark has permitted the processing of distributed data through distributed collections of data (RDDs) using functional transformation. With the growing importance and arena of Spark, it has impacted the wider audiences beyond the field of Big Data and Data Science. In order to assist the modern big data and data science programs and applications, new data frames API was created. This API is very easy to use if new users have some experience with data frames in any other programming language. This API will enable Spark to program *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (91–110) © 2021 Scrivener Publishing LLC

91

92  The Smart Cyber Ecosystem for Sustainable Development it easily. It will also help in improving performance through various methods of codegeneration and intelligent optimizations. Spark SQL named programming module was introduced by Spark for the processing of structured data. It serves as a distributed SQL query engine. Spark SQL architecture comprises of three main layers 1. Language API 2. Schema RDD 3. Data Sources For the manipulation of structured data, we require a domain-specific language that is provided by DataFrame. To explain this further, we also include elaborate examples of processing using DataFrames and operations that can be applied to them. A SQL context helping applications to perform SQL queries automatically in addition to function’s execution and returning of DataFrame. The waiting time associated with queries and running programs is the two main issues related to the processing of large data sets. In this chapter, we will also illustrate techniques for loading and saving data from the various data sources and include in our discussions built-in data sources like JSON Datasets, Hive tables, and parquet files. Chapter Objectives: • • • • • • • •

To understand RDDs, DataFrames, and Datasets To know the use of each set of API Performance and optimization advantages Benefits of DataFrames and Datasets over RDDs To understand the features and importance of Spark SQL To illustrate the architecture of Spark SQL To create SQL data frames and perform data frame operations on them SQL Context helping applications to perform SQL queries automatically in addition to function’s execution and returning of DataFrame • To discuss different types of Data sources like JSON Datasets, Hive tables, and parquet files • To illustrate various methods to load and save data in different Spark Data sources

5.2 Introduction to Apache Spark Apache Spark is an open-source solution for fast computation in cluster computing (field of computing that studies computers connected in a network, distributed systems) [3]. Previously, Apache Hadoop MapReduce was used for this purpose [4]. MapReduce imposed a linear structure of dataflow which increased the latency of applications. So, Apache Spark framework was introduced to extend the MapReduce model for efficient processing of data, execution of queries, and better speed of computations. It uses in-memory cluster computing. Spark covers a great range of application processing such as batch applications, stream processing, execution of iterative algorithms, and applications based on interactive queries.

IoT’s Data Processing Using Spark  93 Due to the integration of various tools together, it allows easy management and in turn increases the processing speed of an application.

5.2.1 Advantages of Apache Spark Increases processing speed: It is observed that Spark is very fast (approximately 100 times) as compared to Hadoop cluster for processing of data at a large scale. This is because of its memory and intermediate optimizations and faster read/write operations to disk. 1. Supports easy-to-use multiple APIs in various programming languages like Java, Python, Scala, and R. 2. A large collection of various operators for transforming data and interactive querying. 3. Allows manipulation of both structured and semi-structured data. 4. Supports advanced libraries for machine learning (ML), graph algorithms, SQL queries, and data streaming. 5. One place for all the tools and standard libraries that increases the productivity of the developer and helps in the creation of complex workflows.

5.2.2 Apache Spark’s Components In this section, we will explain the components of Apache Spark. Figure 5.1 shows the main components of Apache Spark. 1. Apache Spark Core—It is considered a universal execution engine for all platforms supported by Spark. It provides Datasets to reference in external storage systems and In-Memory computing. 2. Spark SQL—This tool furnishes assistance for structured and semi-structured data through SchemaRDD, a new type of abstraction method [5]. 3. Spark Streaming—This is used to perform streaming analytics for fast scheduling. It performs RDD (Resilient Distributed Datasets) transformations on mini-batches of data [6]. 4. MLlib (Machine Learning Library)—This is the dispensed framework of ML to apply algorithms for distributed memory-based Spark architecture [7]. 5. GraphX—This is the distributed graph-processing framework that provides an API for expressing computation and evaluation of graphs [8].

Apache Spark Core (Can be in Scala, Python, R, Java)

Spark SQL

Figure 5.1  Components of Apache Spark.

Spark Streaming

MLlib

GraphX

94  The Smart Cyber Ecosystem for Sustainable Development Map Reduce Map

Compiled Output

Dataset on cloud Map Reduce Map

Figure 5.2  Apache Hadoop MapReduce.

5.3 Apache Hadoop MapReduce Figure 5.2 shows the basic architecture of Apache Hadoop MapReduce. MapReduce is a cluster computing paradigm for generating and processing data sets on a cluster using a parallel, distributed program [9]. Distributed programs have imposed a linear iterative data flow structure by MapReduce [10]. The programs related to MapReduce take a huge amount of data in input from disk. This input is mapped across the data. Then, reduction results are stored on disk after reducing the map’s results. It runs various tasks in parallel, managing communications and data transfers, removing redundancy, and increasing fault tolerance. It is composed of two methods: • Map() method to performs filtering and sorting of the dataset. It converts the dataset into another set of data represented in the form of a key value pair. • Reduce() method takes the output from the map and performs a summarizing operation using analytics of data. It is a simple programming model that focuses on less-cost, flexibility, fault-tolerance, and scalability. The processing of Large Datasets is considered as the main concern associated with the MapReduce.

5.3.1 Limitations of MapReduce 1. Data processing speed—Includes two important tasks: Map and Reduce that require a lot of time to process data which decreasing the speed of processing. 2. Type of data processing—It is designed for processing in the Batch manner and not suitable for data processing in real-time. 3. Latency—Requires a lot of time to perform Map and then Reduce function thereby increasing latency. 4. Ease of use—Need to manually code every operation. It has no interactive mode. 5. Caching—Cannot cache the intermediate data in-memory for a further requirement. 6. Abstraction—Does not have any type of abstraction. 7. Iterative and Graph jobs—Implementing iterative jobs [11, 12] or graph processing is expensive due to the huge space consumption.

IoT’s Data Processing Using Spark  95 Transformation RDD Action

Result

Figure 5.3  Resilient distributed dataset.

5.4 Resilient Distributed Dataset (RDD) To overcome the limitations of MapReduce, the new type of abstraction was created in Spark called RDD (Figure 5.3). The distribution of the collection of items over the nodes of the cluster is known as Resilient Distributed Dataset (RDD) on which transformations and operations are applied in parallel [13]. RDD considered as a multiset of data items that only supports read-only feature partitioned over a clustered or distributed machine continuing in a fault-tolerant way. When to use RDD? • • • •

Processing of streams (unstructured data). When using the functional programming method of manipulating data. When data access and data processing do not depend on schema impositions. When dataset requires low-level transformations and operations.

5.4.1 Features and Limitations of RDDs Following are the features of RDD: 1. Provides efficient and easy processing of various types of data like unstructured and structured data 2. Provides supports for several programming languages like R, Python, Scala, and Java. 3. Stored in Partitions—This allows the processing of huge datasets in parallel using popular distributed algorithms like MapReduce. There is no overhead of work distribution and fault tolerance. 4. Immutability for consistency in computations—Since RDDs are a collection of partitioned records where a partition is the smallest unit of parallel programming, every partition record is immutable, and can be generated with operations on existing partitions. 5. Fault-Tolerant—RDD allows redoing transformations on the same partition when some data is lost. This helps to achieve the same computation results again without replication of data across multiple nodes. 6. Lazy Evaluations—All transformations are said to be lazy because the computation of each transformation is done when required that is when asked by the caller program.

96  The Smart Cyber Ecosystem for Sustainable Development HDFS File

Filtered RDD

Mapped RDD

Figure 5.4  Spark RDD operations—transformation and action.

7. Functional operations—Mainly two kinds of operations known as transformations like map(), filter(), and reduce(), and actions such as collect() and saveAsObjectFile() supported by RDD. Creating a new dataset from the already existed one is done using Transformation. Actions return computed value to the calling program obtained by processing the datasets. Figure 5.4 depicts the operations of Spark RDD. Following are the limitations of RDDs: 1. Spark does not provide support for built-in optimization for RDDs. Each RDD is optimized by the developer based on its attributes. 2. RDDs require the user to specify the schema explicitly because it does not infer it on its own.

5.5 DataFrames DataFrames overcomes the limitations of RDDs as the data is organized into a relational manner. It provides metadata for the distributed data that is equivalent to tables in a relational database. For efficient and easier processing, they superimpose a structure on the distributed collection of data. Building an extensible query optimizer, Spark acts as a catalyst optimizer. When to use DataFrames? • When lambda functions and columnar access is needed to be performed on semi-structured data. • When low-level functionality is required. • When processing high-level operations like map, reduce, filter, aggregation, etc. • When the usage of domain-specific APIs is required. Features of DataFrames: 1. Distributed collection of Row Object: This can be seen as a table in the relational database but equipped with optimization methods. 2. Several data formats are supported: Processing storage systems (MYSQL, HDFS, HIVE tables, etc.), structured, and unstructured data formats (Cassandra, elastic search, CSV, and Avro). 3. Read and Write from different data sources. 4. Use of catalyst optimizer for optimization: Along with DataFrame APIs, catalyst optimization also supports optimization for SQL queries. Optimization using catalyst tree transformation is done in four phases.

IoT’s Data Processing Using Spark  97 5. SQL Capabilities - Supports Spark SQL and gives complete compatibility with existing data, UDFs, and queries of Hive. 6. It provides supports for several programming languages like R, Python, Scala, and Java. Limitations of DataFrames: 1. Compile-time safety—DataFrame API does not offer compile-time safety. This makes manipulation of data(transformation and aggregation) difficult when the data structure is not known. 2. Cannot regenerate domain Object—Domain object cannot be regenerated and operated upon after it is transformed into a DataFrame.

5.6 Datasets Dataset API can be described as an addition to DataFrame API that imparts a runtime as well as a compile-time type-safe, object-oriented programming interface. Datasets contains two discrete APIs characteristics—strongly typed and untyped. A DataFrame can be seen as a group of generic type Dataset[Row], where the Row is a general untyped JVM object whereas Datasets are a collection of strongly typed JVM objects. Datasets are a collection of objects that are immutable and strongly-typed mapped to some relational schema. At the core of the API is an encoder that converts JVM objects into tabular representation. This is done to improve memory utilization. The types of discrete APIs characteristics in datasets can be seen in Figure 5.5. When to use Datasets? • • • • •

When high-degree runtime safety is required. When we need to use typed JVM objects. When we use Catalyst optimizer. When we need to save memory. When we require fast computations.

Features of Datasets: 1. Type Safety—Datasets API provides runtime-type and static-type safety which was not available in DataFrames.

RDD DataFrame

Datasets

Untyped API (DataFrame = Dataset[Row]) Typed API (Dataset[T])

Figure 5.5  Types of discrete APIs characteristics in datasets.

98  The Smart Cyber Ecosystem for Sustainable Development 2. Combines the best features of RDD and DataFrame API. These include features like type-safe, functional programming from RDD API as well as features like sorting and shuffling, Tungsten execution, Query Optimization, and the relational model. 3. Encoders—Any JVM object can be converted into a Dataset. This allows customers to work with both structured, semi-structured, and unstructured data. 4. This functionality was not available in the DataFrame API. 5. Interoperable—Existing RDDs and DataFrames can be easily converted into Datasets. Limitations of Datasets: 1. Data is queried from Datasets which further requires specifying the fields in the class typecast as a string. After the data is queried, the column is cast to the required data type. 2. Datasets only support Scala, Python (Spark 2.0 onward), and Java. It does not support R. Performance Comparison of three APIs: 1. 2.

Comparison based on type safety Datasets provide compile-time as well as runtime type safety. Comparison based on Performance Optimization The DataFrame and Dataset APIs use Catalyst to generate optimized logical and physical code in various languages like Java, Scala, or Python. Also, it is observed that the Dataset [T]-type API is optimized for engineering tasks and the DataFrame API performs interactive analysis faster. 3. Comparison based on Space Optimization Datasets use very less memory cache (less than 15 GB) when compared to RDD (approximately 60 GB).This is because it generates compact bytecode due to the presence of encoders in the Dataset API that efficiently serializes and de-serializes JVM objects. In summary, it is very obvious choosing the use of RDD or DataFrame and/or Dataset [14]. Dataset focuses on custom structure and view and offers domain-related and high-level functionalities. It saves space and supports very fast execution speed. RDD or DataFrame provides low-level control and functionalities.

5.7 Introduction to Spark SQL Spark SQL is a component of Spark Core that is used to work with structured, semi-structured, and unstructured data [15]. The data having some kind of schema (set of known fields for each record) such as JSON, Hive tables, and Parquet known as Structured data. The Semi-Structured data proposes no difference between the data and schema. The Spark

IoT’s Data Processing Using Spark  99 SQL helps in providing a domain-specific language (DSL) to manipulate a data abstraction element known as DataFrames. Spark SQL supports SQL language having a command-line interface and ODBC/JDBC server. To use Spark SQL, we need to import its package as follows: import org.apache.spark.sql We will be using Scala for our examples further in this chapter. Capabilities of Spark SQL: 1. Spark SQL helps in providing DataFrame abstraction in various programming languages like Java, Python, and Scala to go with structured data. 2. Helping in reading and writing of data in various structured formats supported by JSON, Parquet files, Hive tables, etc. 3. Spark SQL helps you in querying the data and supports internal and external tools through standard DB connectors (JDBC/ODBC). Features of Spark SQL: 1. Integrated—Spark programs support the use of Spark SQL. Spark having query structured data as a distributed dataset (RDD) is supported by Spark SQL integration. 2. Unified Data Access—This allows loading and querying of information from various sources. Schema-RDDs in Spark can load data from parquet files, Apache Hive tables, and JSON files. 3. Hive Compatibility—This helps running of unmodified Hive queries on existing warehouses. We can use existing Hive data, queries, and user-defined functions thus, reusing Hive frontend. 4. Standard Connectivity—It can be connected with industry-standard JDBC or ODBC. 5. Scalability—The same engine is used by Spark SQL for both interactive and long queries. This helps in scaling large jobs and taking advantage of the RDD model for supporting mid-query fault tolerance.

5.7.1 Spark SQL Architecture Three layers of Spark SQL architecture are (Figure 5.6): 1. Language API—Various programming languages like Python, Scala, and Java support and compatible with Language API. 2. Schema RDD—The schemas like tables and records are used and maintained by Spark SQL. It uses the Schema RDD also called Dataframe as a temporary table. 3. Data sources—A data source for spark SQL are JSON document, HIVE tables, Cassandra DB, and Parquet file.

100  The Smart Cyber Ecosystem for Sustainable Development Language API (Scala, Python, Java) Schema RDD (Dataframe) Data Sources for Spark SQL (Parquet files, JSON files, HIVE tables)

Figure 5.6  Spark SQL architecture.

5.7.2 Spark SQL Libraries To perform relational and procedural processing, Spark SQL provides the following libraries: 1. Data Source API (Application Programming Interface) Data Source API is a general API used for loading and storing structured data. The various data source like Avro, Hive, JSON, JDBC, Parquet, etc., are supported by in-built features of Data Source API. The Spark packages and smart sources of API help in integration with the third party. 2. DataFrame API This API contains row and column as a distributed collection of data. The SQL context and Hive Context help in retrieving this API. It is evaluated similarly to the lazy behavior of Apache Spark Transformations. The data formats like Cassandra, CSV, Elastic Search, etc., are being supported by different data formats. Various programming languages like Python, Scala, Java, and R supports DataFrame API and can be easy with BD (Big Data) tools and frameworks using Spark-Core. 3. SQL Interpreter And Optimizer It is constructed in Scala supported by functional programming. This component is considered as most technically evolved and newest. The transforming trees are used for planning, evaluation and analysis, and optimization, etc. Cost-based and rule-based optimization is supported by SQL interpreter and optimizer. This is much faster than their counterpart’s RDD in terms of the execution of queries. Each rule in the framework concentrates on distinct optimization. 4. SQL Service SQL service acts as an entry point for structured data in Spark and helping the execution of SQL queries and the formation of DataFrame objects.

5.8 SQL Context Class in Spark The functionalities of Spark SQL are initialized with the help of SQL Context class. With the starting of spark-shell, SparkContext object named as sc is initialized by default. SQLContext class or its related descendants act as an entry point into all functionality in

IoT’s Data Processing Using Spark  101 Spark SQL. The commands used to initialize SQLContext with the help of spark-shell are shown below: val sc: SparkContext val sqlContext = new org.apache.spark.sql.SQLContext(sc) Converting an RDD to a DataFrame using SQLContext class implicitly: import sqlContext.implicits._ SQLContext having a sql() method is called to make a query against a table. Example -> We are loading some data regarding students from JSON. It provides a name registering as “Students” and further querying it with SQL. Our file students .JSON contains: [{"RNo" : 1, "name" : "Rohit", "age" : 15, "weight": 40},  {"RNo" : 2, "name" : "Rahul", "age" : 14, "weight":58},  {"RNo" : 3, "name" : "Raj", "age" : 15, "weight": 50},  {"RNo" : 4, "name" : "Rishab", "age" : 16, "weight": 47},  {"RNo" : 5, "name" : "Ramit", "age" : 15, "weight": 49}] Code to query data: objectBasicQueryExample { valsc = SparkCommon.sparkContext valsqlContext = new org.apache.spark.sql.SQLContext(sc) def main(args: Array[String]) { importsqlContext.implicits._ val input = sqlContext.read.json("src/main/resources/students.json") input.registerTempTable("Students") val result = sqlContext.sql("SELECT * FROM Students") result.show() }}

5.9 Creating DataFrames We can create DataFrames in different ways [16]: 1. DataFrames can be easily developed with various data formats. For example, loading of source data from JSON and CSV files. 2. It can be created by loading data from Existing RDD. 3. It can be created by programmatically specifying the schema. Different applications can be created with DataFrames using existing RDD with the help of SQLContext. This application uses data from the Hive table or some other data sources. The below code shows the creation of DataFrame based on the content of students.

102  The Smart Cyber Ecosystem for Sustainable Development As an example, the following creates a DataFrame based on the content of students.JSON file that is read using the read.json() command: val sc: SparkContext val sqlContext = new org.apache.spark.sql.SQLContext(sc) val df = sqlContext.read.json(“examples/src/main/resources/students.json”) To displays the contents of the DataFrame to standard output use: df.show()

5.9.1 Operations on DataFrames The language that is domain-specific is provided by DataFrame for the manipulation of structured data. Here, we include some basic operations on structured data processing using DataFrames. To show an example, we are converting student.json file to data frame. 1. show()—To see first 20 rows of DataFrame in a tabular form studentDataframe.show() 2. show(n)—To see first n rows of DataFrame in a tabular form studentDataframe.show(2) // shows 2 rows 3. printSchema()—To see the schema or structure of the DataFrame studentDataframe.printSchema() 4. select()—To select a particular column among all columns from the DataFrame. studentDataFrame.select(“age”).show() 5. take(n)—To return the first n rows studentDataFrame.take(3).foreach(println) 6. count() –To return the number of rows. studentDataFrame.groupBy(“age”).count().show() 7. head()—To return the first row. valresultHead = studentDataFrame.head() println(resultHead.mkString(“,”)) 8. head(n)—To return first n rows. val resultHeadNo = studentDataFrame.head(5) println(resultHeadNo.mkString(“,”)) 9. first()—Another command to return the first row. val resultFirst = studentDataFrame.first() println(“first:” + resultFirst.mkString(“,”)) 10. toDF()—It will return a renamed coloumns in a new DataFrame. This function is used in conversion to DataFrame from a RDD with meaningful names. val student = sc.textFile(“src/main/resources/students.txt”) .map(_.split(“,”)) .map(f => Student (f(0).trim.toInt, f(1), f(2).trim.toInt)) .toDF().show()

IoT’s Data Processing Using Spark  103 11. dtypes()–To return all column names and their data types as an array. studentDataFrame.dtypes.foreach(println) 12. columns()—To return all column names as an array. studentDataFrame.columns.foreach(println) 13. sort()—To return a new DataFrame sorted by the given expressions. studentDataFrame.sort($”RNo”.desc).show() 14. orderBy()—To return a new DataFrame sorted by the specified column(s). studentDataFrame.orderBy(desc(“age”)).show() 15. groupBy()—To group by a column studentDataFrame.groupBy(“age”).count().show() 16. filter()—To filter by athe given expression studentDataFrame.filter(studentDataFrame(“age”) >15).show() 17. where()—To filter using the given SQL expression. studentDataFrame.where($”age” >15).show() 18. unionAll()—This gives a new DataFrame having combination of rows in both frames. studentDataFrame.unionAll(tempDataFrame).show() 19. intersect()—This gives a new DataFrame having row only in both frames. studentDataFrame.intersect(tempDataFrame).show() 20. except()—This gives a new DataFrame having combination of rows in one frame but in another frame. studentDataFrame.except(tempDataFrame).show() 21. drop()—To return a new DataFrame with a column dropped. studentDataFrame.drop(“age”).show()

5.10 Aggregations The common aggregations functions like count(), countDistinct(), avg(), max(), min(), etc., are some of the in-built functions provided by DataFrames. Scala and Java support typesafe versions provided by Spark SQL to work with strongly typed datasets. Often, users are not confined to the in-built aggregate functions and can create their own. Aggregate functions usually return a unique single value. If we want to find the aggregate values for each unique value in a column, we should usegroupByover that column to build the groups. There are two types of aggregation functions: 1. Untyped User-Defined Aggregate Functions—This type of function works with zero or more rows to produce one row. 2. Type-Safe User-Defined Aggregate Functions—This type of function may produce more than one row of the result.

5.11 Running SQL Queries on DataFrames An SQLContext allows various applications to execute SQL queries automatically in addition to the execution of SQL functions and returning results as a DataFrame. Apart from this, the Spark session provides similar functionalities. Temporary views are session-scoped

104  The Smart Cyber Ecosystem for Sustainable Development and will vanish as soon as the session terminates in Spark SQL. We can create a global temporary view that would help us in creating a view that is shared among all the present sessions and keep the view alive till Spark application ends.

5.12 Integration With RDDs We can convert existing RDDs into Datasets with the help of two methods in Spark SQL. In the first method, the schema is deduced with the support of RDD that accommodates the specific types of reflection and objects. This method encourages compact code and efficient results when the schema is already known. In the second method, Datasets are created using a programmatic interface by schema’s construction, and then existing RDD is applied. Two different methods are supported by Spark SQL for converting existing RDDs into DataFrames: 1. Inferring the schema using reflection—Generation of the schema of an RDD using reflection that accommodates specific types of objects. 2. Programmatically specifying the schema—Using a programmatic interface to build the schema and it is applied to the existing RDD.

5.12.1 Inferring the Schema Using Reflection Spark SQL’s scala interface supports automatic conversion of an RDD to a DataFrame when RDD contains case classes i.e., those classes that define the table schema. Using reflection, case class-related arguments are read and the names of the columns are set corresponding to the names of these arguments. The complex types such as arrays are nested and accommodated by Schema. DataFrame can be obtained by converting RDD and then registered as a table. The SQL statements use this table.

5.12.2 Specifying the Schema Programmatically In the second method, we use the programmatic interface to construct the schema and to create DataFrame, and it is applied to existing RDD. As discussed, this method is useful when case classes are not defined ahead of time, that is, they are defined at run time. The following three steps are used to create DataFrame programmatically: 1. Using original RDD create an RDD of rows. 2. Schema is created which is represented by Struct Type duplicating the structure RDD’s rows created in Step 1. 3. Now, through the createDataFrame method inSparkSession, this schema is applied to RDD.

5.13 Data Sources The DataFrame interface and Spark SQL allow data input from various data sources. A Data Source is defined as a temporary table and behaves like a normal RDD. This approach of

IoT’s Data Processing Using Spark  105 SparkQL

HiveQL Hive queries Hive parser

Spark SQL queries

Dataframe DSL

SparkSQL Parser

DataFrame

Catalyst

Spark RDO code

Figure 5.7  Spark SQL pipeline.

registering as a table is helpful in running SQL queries over its data. Spark SQL pipeline can be viewed in Figure 5.7. In this chapter, we will describe the following types of data sources:

5.13.1 JSON Datasets Spark SQL can easily contain the JSON dataset’s schema and load it into it as a DataFrame. Each line in this JSON file accommodates a separate self-contained valid JSON object and not similar to the traditional JSON file. It must be separated using a newline-delimited JSON. The conversion of this dataset can be done using SQLContext.read.json(). We can directly query JSON data using spark SQL. Also, we can automatically infer JSON schemas for both reading and writing data and loads it as a Dataset[Row]. Users can directly access nested fields in JSON data through Spark SQL without any explicit transformations.

5.13.2 Hive Tables Hive produces a huge number of functionalities that are dependent [17]. Spark loads them automatically if they are found on the class path. If all the worker nodes contain the Hive dependencies, then it is required to get access to Hive serialization and deserialization libraries (SerDes) for accessing the data stored in Hive. To create a Hive table, we need to define how this table will read (Output format) and write (Input format) data from/to file system. It is necessary to define how to deserialize the data to row and vice-versa in the table. The table files are read as plain text by default. Figure 5.8 shows the architecture of Hive tables.

HiveQL Metastore

UDFs Spark SQL Apache Spark

Figure 5.8  Architecture of Hive tables.

SerDes

106  The Smart Cyber Ecosystem for Sustainable Development

5.13.3 Parquet Files Parquet files are written in columnar format. We can read and write these files such that the schema of the original data is automatically preserved when we use spark SQL. For compatibility, all the columns are converted to nullable while writing Parquet files. The advantages of having columnar storage are as follows: • • • •

Limiting I/O operations. Fetching the defined columns requiring access. Consuming very less space. Giving summarized data in a better manner following the type-specific encoding.

Both reading and writing of Parquet files are supported by Spark SQL. It follows the same method of reading and writing as JSON Datasets. Also, table partitioning like in Hive is used as an optimization technique. The Parquet files can capture the partitioning information. Normally, data are stored in different directories in the partitioned table.

5.14 Operations on Data Sources This section describes different operations: 1. Load/Save Functions We can configure which data source to use by spark.sql.sources.default. The default data source is parquet file which will be used for all operations demonstrated here. val studentDataFrame = spark.read.load("examples/src/main/resources/student .parquet ") studentDataFrame.select("name", "age").write.save("namesAndAge.parquet") 2. Specifying Options for Data Sources We can specify different options manually along with the data source that will be used. Data sources can be specified using two methods: 1. Using the fully qualified name (i.e., using org.apache.spark.sql.parquet). 2. Using the short names (i.e., JSON, parquet, JDBC, orc, libsvm, CSV, text). Also, using the below syntax, we can convert them to different types: valstudentDataframe = spark.read.format("JSON").load("examples/src/main/ resources/students.json") studentDataframe.select("name", "weight").write.format("parquet").save("names AndWeight.parquet") 3. Using SQL directly We can directly query the file using SQL instead of using read API to load the file into DataFrame and then query it.

IoT’s Data Processing Using Spark  107

val sqlDF = spark.sql("SELECT * FROM parquet.`examples/src/main/resources/ student.parquet`")

4. Modes of saving The SaveMode is optionally available for save operations to specify how to handle existing data. These save modes are not atomic and do not utilize any locking. Also, data is deleted before performing an Overwrite. • • • •

SaveMode.ErrorIfExists (default) SaveMode.Append SaveMode.Overwrite SaveMode.Ignore

5. Bucketing, Sorting, and Partitioning Different operations are available for the output of file-based data source— bucket, sort, and partition. Persistent tables support bucketing and sorting and can be done as follows: studentDataFrame.write.bucketBy("name").sortBy("age").saveAsTable ("student_bucketed") Partitioning can be used with both save and saveAsTable when using the Dataset APIs and is used as follows: studentDataFrame.write.partitionBy("age").format("parquet").save("namesPartByAge.parquet")

Both partitioning and bucketing can be used for a single table as shown:

studentDataFrame.write.partitionBy("age").bucketBy("name").saveAsTable ("student_partitioned_bucketed")

Directory structures are created by partitionBymethod having limited applicability to columns having high cardinality. In contrast, data is distributed across the fixed number of buckets in the bucketBy while having unbounded unique values.

5.15 Industrial Applications The different survey has been conducted by founding members of Apache Spark on “Why companies should use in-memory computing framework like Apache Spark?” and some key points are described below. Below are some of the use cases of spark in industry demonstrating the ability of spark to build and run the big data applications faster. 1. Finance Industry 2. e-commerce Industry

108  The Smart Cyber Ecosystem for Sustainable Development 3. Healthcare Industry 4. Media and Entertainment Industry 5. Travel Industry

5.16 Conclusion So, in the first section, we introduce Apache Spark and how it has taken over and improved Hadoop MapReduce. We see that MapReduce uses the linear flow of structure which restricts the speed of processing computations in cluster programming and does not support graph and iterative procedures. So, to overcome these disadvantages, Spark was introduced. It can be used along with MapReduce to improve its performance as well as can act as an independent processing engine to process data in distributed computing. Spark covers a great range of application processing such as batch applications, stream processing, execution of iterative algorithms, and applications based on interactive queries. It increases the speed of computations and supports multiple APIs in different languages. We also see the components of Spark—Spark Core, Spark SQL, Spark streaming, ML libraries, and Graph processing component. These components increase the range of applications addressed by Apache Spark. Then, we discuss three different APIs offered by Spark, that is, RDDs, DataFrames, and Datasets. We discuss where each of these APIs is used and their advantages and limitations. In summary, low-level functionalities and control are offered by RDD and DataFrame/Dataset permits custom view and offers high level and domain-specific operations saving hard disk memory and making execution at very high speed. After this, we introduce the Spark SQL component and discuss its features and capabilities. Then, we understand the architecture of Spark SQL and the libraries provided by it. We discuss SQLContext that is used for initializing the functionalities of Spark SQL. It can also be done using SparkSession. We understand procedures to create data frames, operations that can be performed on them, and how to run SQL queries to query data from data frames. Also, we discuss three types of data sources available to input data and operations that can be performed on them. In summation, it allows simplification of challenges and huge computing tasks involving high volumes of real-time and archived data (structured and semi-structured data). It helps in integrating related complex capabilities like ML and graph algorithms.

References 1. Alexandrov, A., The Stratosphere Platform for Big Data Analytics. VLDB J., 23, 6, 939–964, 2014. 2. Ko, S.Y., Hoque, I., Cho, B., Gupta, I., On Availability of Intermediate Data in Cloud Computations. Proceedings of the 12th Conference on Hot Topics in Operating Systems, USENIX Association, Berkeley, CA, USA, pp. 6–6, 2009. 3. Apache Spark project. http://spark.apache.org. 4. Dean, J. and Ghemawat, S., MapReduce: Simplified Data Processing on Large Clusters. Proceedings of Operating Systems Design and Implementation, San Francisco, CA, pp. 137–150, 2004.

IoT’s Data Processing Using Spark  109 5. Spark, S.Q.L., DataFrames, and Datasets Guide. http://spark.apache.org/docs/latest/sqlprogramming-guide.html 6. He, B., Yang, M., Guo, Z., Chen, R., Su, B., Lin, W., Zhou, L., Comet: Batched Stream Processing for Data-Intensive Distributed Computing. In Proceedings of the 1st ACM Symposium on Cloud Computing, Indianapolis, USA, pp. 63–74, 2010. 7. MLlib user guide. https://spark.apache.org/docs/latest/mllib-guide.html 8. Malewicz, G., Austern, M.H., Bik, A.J., Dehnert, J.C., Horn, I., Leiser, N., Czajkowski, G., Pregel: A System for Large-Scale Graph Processing. Proceedings of the ACM SIGMOD International Conference on Management of Data, Indianapolis, USA, pp. 135–146, 2010. 9. He, Y., Lee, R., Zheng, S., Jain, N., Xu, Z., Zhang, X., RCFile: A Fast and Space-Efficient Data Placement Structure in MapReduce-Based Warehouse Systems. IEEE International Conference on Data Engineering, Hanover, Germany, pp. 1199–1208, 2011. 10. Bu, Y., Howe, B., Balazinska, M., Ernst., M.D., HaLoop: Efficient Iterative Data Processing on Large Clusters. Proceedings of the VLDB Endowment, vol. 3, no. 1, pp. 285–296, 2010. 11. Bhatotia, P., Wieder, A., Rodrigues, R., Acar, U.A., Pasquin, P., Incoop: MapReduce for Incremental Computations. Proceedings of the 2nd ACM Symposium of Cloud Computing, Cascais, Portugal, pp. 1–14, 2011. 12. Moretti, C., Bui, H., Hollingsworth, K., Rich, B., Flynn, P., Thain, D., All-Pairs: An Abstraction for Data-Intensive Computing on Campus Grids. IEEE Trans. Parallel Distrib. Syst., 22, 1, 33–46, 2010. 13. Zaharia, M., Chowdhury, M., Das, T., Dave, A., Ma, J., McCauley, M., Franklin, M., Shenker, S., Stoica, I., Resilient Distributed Datasets: A Fault-Tolerant Abstraction for In-Memory Cluster Computing. Proceedings of the 9th USENIX Conference on Networked Systems Design and Implementation, San Jose, CA, pp. 2–2, 2012. 14. Pavlo, A., A comparison of approaches to large-scale data analysis. Proceedings of the ACM SIGMOD International Conference on Management of Data, Rhode Island, USA, pp. 165–178, 2009. 15. Tahara, D., Diamond, T., Abadi, D.J., Sinew: A SQL System for Multi-Structured Data. Proceedings of the ACM SIGMOD International Conference on Management of Data, Utah, USA, pp. 815–826, 2014. 16. Konrath, M., Gottron, T., Staab, S., Scherp, A., SchemEX - Efficient construction of a data catalog by stream-based indexing of linked data, in: Web Semantics: Science, Services, and Agents on the World Wide, vol. 16, no. 5, pp. 52–58, 2012. 17. Thusoo, A., Hive–a petabyte-scale data warehouse using Hadoop. Proceedings of the 26th International Conference on Data Engineering, California, USA, pp. 1–6, 2010.

6 SE-TEM: Simple and Efficient Trust Evaluation Model for WSNs Tayyab Khan and Karan Singh* SC&SS, JNU New Delhi, India

Abstract

Trust estimation approaches (TEAs) in wireless sensor networks (WSNs) are mainly used to improve security, reliability, system efficiency, and cooperation among sensor nodes (SNs). There are various existing TEAs unable to satisfy the fundamental requirements of WSN, like resource efficiency, communication, and memory overheads. In this paper, we provide an efficient TEA to alleviate several internal attacks (threats) like badmouthing, black hole, and gray-hole attacks for clustered WSN. The proposed method improves cooperation (dependability) and provides adaptability and flexibility in terms of application requirements with reduced overhead. Experimental results exhibit great performance in terms of severity analysis, attack detection, prevention, and mitigation of selfish (malicious and bad) nodes to protect WSN. Keywords:  Trust, wireless sensor network, trust estimation, direct and indirect trust, intra-cluster, inter-cluster

6.1 Introduction Wireless sensor network (WSN) consists of numerous tiny-size SNs to sense events like environmental or battlefield monitoring and report data. Each sensor node (SN) is made up of processor, memory, battery transceiver, and sensing sub-system. The sensed data is processed and transmitted to the base station (BS). The BS is also known as sink node that has enough resources to perform tasks. The BS acts as an interface between WSN and its users. Users can extract the required relevant information from the BS by injecting appropriate queries. A router plays a vital role in gaining reliability in addition to long-distance communication between SNs and gateways [1, 2]. SNs communicate their sensor readings with neighbor nodes within communication range by sending messages via relay nodes. The communication among SNs may be single-hop or multi-hop. During single-hop communication, a SN can directly exchange messages with other SNs within the communication range. Although, in large WSNs, it is infeasible to direct exchange messages with each other SNs. These messages are also known as data packets [2]. Multi-hop communication provides the facility to transmit and exchange data packets with every SN in the WSN through *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (111–130) © 2021 Scrivener Publishing LLC

111

112  The Smart Cyber Ecosystem for Sustainable Development relay nodes. Multi-hop communication relates two different domains to send out a message from source to BS through a sequence of pair wise adjacent SNs. Most of the WSNs nodes communicate in multi-hop fashion via radio links with limited available bandwidth and form a temporary network. This temporary network is a network without predefined infrastructure as well as without centralized network administration [3]. WSN uses a highly dynamic network topology where, any time, SNs can leave-joins a network and change their locations. SNs work in two modes, namely, continuous mode as well as event-driven mode. A SN with working mode along with GPS and local positioning algorithms plays an essential role in obtaining information related to location and position, respectively. WSNs are being used in different domains as well as found suitable for a wide-ranging real-time applications due to its sensing, processing, along with decision-making capability. Due to the unattended mass deployment nature, no global IDs, and highly dynamic network topology of WSNs, SNs are less reliable, failure-prone, and prone to insider or outsider security attacks such as an on-off attack, Sybil attack, DoS, etc. [4, 5]. Due to these attacks, a node becomes malicious in the form of untrustworthy, non-reliable, or selfish. Thus, security is an essential issue in WSN to preserve confidentiality, availability, and integrity of sensed data. Cryptographic techniques like authentication, encryptions are competent against external threats as well as not able to catch malicious behavior of insider adversarial nodes [6]. Figure 6.1 shows the components of SNs and WSN’s architecture.

Sensor node passing sensed data

Internet and Satellite C Sink

E

D

B A

Task Manager Node

Sensor field

User

Sensor Nodes Location Finding System

Sensing Unit Sensor

ADC

Mobilizer

Processing Unit Processor Storage

Transceiver

Power Unit

Figure 6.1  Components of sensor nodes and WSN’s architecture.

Power Generator

SE-TEM for WSNs  113

6.1.1 Components of WSNs The usual infrastructure of WSN consists of various components [7, 8] like SNs, controller, gateways, and task manager, as shown in Figure 6.2. The SNs act as source or sink. This subsection discusses the components of a typical WSN. Sensing Unit: It consists of ubiquitous sensors to monitor and control environmental conditions as well as provide a real interface to the physical world. These SNs collect sensitive information from the targeted surroundings. The sensing unit is further attached with analog-to-digital converter (ADC) that converts an analog signal to the digital signal. Now, this digital signal is appropriate for the microprocessor. Controller: It is the CPU of SNs that process as well as analyzed the gathered data and made appropriate decisions. Power Unit: It is the most essential and critical system component of WSNs for the proper functioning of the network system. Without power in SNs, WSN is no longer beneficial. Moreover, sufficient energy is required for the proper functioning of the processing unit as well as the transmitting unit. Planting a solar cell is one of the available solutions to generate a tremendous amount of power. Transmitting Unit: It is used to transmit the signal towards the sink node or neighbors SNs with short-range communication. In WSNs, a SN can be in “transmit mode, sleep mode, idle mode, or receive mode.” It is beneficial to keep a node in sleep mode whenever it is not sensing, transmitting or receiving any sensitive information to reduce energy consumption. Task Manager: It can be visualized as a platform for information retrieval, processing, and analysis by employing any display interface since it stores all data and raw information generated by the SNs. Memory: It is used to store transitional programs, intermediate sensor readings, and data required for efficient decision-making. Communication Device: It is responsible for exchanging (sending or receiving) data between SNs. Gateway: Gateways provide an interface between SNs and systems (PC and PDAs) as well as, sometimes, act as a proxy. WSN is an emerging and promising monitoring technique for several futuristic applications such as environmental, defense, water, health, combat monitoring, home, industries, habitat monitoring, agriculture monitoring, and intruder detection since it provides

Memory

Communication Device

Controller

Power Supply

Figure 6.2  Components of WSNs.

Sensors

114  The Smart Cyber Ecosystem for Sustainable Development real-time monitoring using sensors along with wireless communication and processing power. In recent years, sensor networks have gained worldwide attention from the researchers since it is promising remote monitoring technique. However, in several applications and hostile environment, some safety applications are necessarily required to protect the sensor network from severe security threats. There are various challenges and constraints [15] involved in designing and implementation of a secure wireless communication system that is listed as follows. Security and Privacy: Wireless channels are accessible to both legitimate and illegitimate nodes within the specified communication range. Due to open and remotely accessible environments, WSNs are exposed to numerous internal attacks, including wormhole attack, replay attack, selective forwarding, garnished attack, rushing attack, gray hole attack, and impersonation attack that results in compromised data confidentiality and node authenticity. An adversary can destroy critical information, misguide about shorter routing paths, time synchronization, localization, and several other network services by intercepting sensitive information. Protection from the above-mentioned severe attack is the crucial and challenging difficulty that a WSN faces these days. Hence, achieving acceptable security with constrained resources is a primary issue in WSNs that needs to be addressed. Although various security solutions, including node authentication, trust modeling, key establishment, and distribution, have been proposed, still some improvement is required in terms of robustness, scalability, convergence, and resource efficiency. Limited Memory and Storage: The memory of the sensor ranges from 2 KB to 256 KB, while storage ranges to 32 KB to 2 GB. For security purposes, highly efficient algorithm implementation is required, which leads to more memory consumption, limited memory, and storage makes it difficult. Limited Power: Energy-efficiency is the most prominent issue as well as a design challenge in WSN. SNs are dependent on battery for energy; therefore, energy consumption must be minimized or optimized for long life. The lifetime of SNs tightly depends on the battery life since the battery is consumed at the time of sensing, communication as well as data processing. Unreliability of Communication: The wireless medium is open and accessible to anyone. Due to this, any transmission can easily be intercepted and altered. Moreover, packet loss may take place due to unreliable transmission channels. Limited Bandwidth and Low Data Rates: The accessible bandwidth and data rates (10– 100 Kbits/second) of the WSN’s channel are limited as well as depend on transmission range. Since sensor networks operate in bandwidth-limited, open, unattended as well as a multi-hop communication medium, synchronization among message exchanges is difficult to achieve due to this limited bandwidth. Deployment and Immense Scale: Frequent topology changes due to node mobility in WSNs caused node failures and environmental obstructions. The SN can be deployed in large areas; it is therefore required that the efficient security mechanism can operate within this dynamic environment. Operation unattended: SNs may be left unattended for a long period of time, which results in the SNs face the destruction or capture and compromise by attackers. Routing: The layered architecture of WSNs makes it vulnerable to various security threats, as mentioned above, including routing attacks. Although several solutions have been proposed so far to make secure routing, the collaboration of SNs for routing is still

SE-TEM for WSNs  115 vulnerable to routing attacks. Attackers may inject false/wrong information in the routing paths, redirect traffic, launch DoS attack, and propagate false information in the sensor network. Hence, designing a reliable as well as energy-saving routing protocol is a severe challenge in WSNs. Transmission Error: The performance of WSN is primarily affected by numerous factors such as heat, temperature, transmission errors, bandwidth, data rate, storage, and processing capability. These factors cause high packet loss, delay variance, and high bit error rate. Hence, designing an efficient system that can reduce transmission errors is itself a significant challenge in WSNs. Cost and Size: SNs should be small and low-cost, which results in great acceptability at a worldwide scale as well as incurs low cost and size of synchronization algorithm. A smallsized and inexpensive SN reduces the overall financial metric of WSNs, and more number of users can adapt it to fulfill their requirements. Design Constraints: The primary and often most essential design challenge for a WSN is to build a tiny, inexpensive as well as efficient sensor device. Moreover, additional challenges on hardware and software models with various limitations can affect the design of SNs. Quality of Service: In real-time applications such as remote healthcare, the sensor must provide a high quality of service for accurate analysis as well as decision-making. However, sensor networks must provide an acceptable level (lower bound) of service to its user(s) that are a challenging task under mobility (dynamic topology). Moreover, moderate bandwidth is required to achieve a minimal acceptable quality of service. Various factors affect QoS, such as unbalanced traffic, data aggregation, multi-hop routing, and varying network density. Self-Management: WSNs should be able to do self-management in terms of network configuration, repair, maintenance, and adaptation without any human intervention. Scalability: A WSN should be scalable since any time SNs need to be injected or removed depending upon the application scenario. Moreover, a synchronization scheme should be flexible with the varying density of the sensor network. Robustness: The sensor network should be robust against the misbehavior of faulty SNs and maintain normal workflow without affecting other SNs along with synchronization schemes whenever any SN(s) goes down.

6.1.2 Trust Security is an essential issue in WSN to preserve confidentiality, availability, and integrity of sensed and transmitted data. Cryptographic techniques like authentication, encryption are used to provide security against external attacks and not able to catch malicious behavior of insider adversarial nodes (compromised nodes) [1]. In such cases when malicious (untrustworthy, selfish, faulty, or bad) SNs misbehave, trust estimation approaches (TEAs) are the only solution to protect it against internal threats like rushing attack, black hole, gray hole, Sybil attack and on-off attacks for the survival of WSN. The concept of trust is originated from human society and hence plays an essential role in our everyday lives. Trust is defined as a relationship between trustor and trustee. Trust describes a subjective relation among entities, while reputation is an opinion about an entity. These opinions are provided by participating entities in a relationship. Hence, trust

116  The Smart Cyber Ecosystem for Sustainable Development may be used to determine the reputation of an entity, and reputation is used to determine the trustworthiness of an entity [6]. Trust specifies the reliability or trustworthiness of an entity. Though several definitions of trust are available in the existing literature, the most cited definition of trust is presented by Dasgupta [7, 8] as “the expectation of one person about the actions of others that affect the first person’s choice, when an action must be taken before the actions of others are known”. However, other definitions of trust are listed as follows. According to [9], trust is the “reputation of the entity where reputation is the opinion about others”. Trust is a belief [45] that ensures the entity as secure and reliable. In [10], trust is defined as how much an entity matches the expectations of another entity. Trust is subject matter specific, i.e., related to the particular function, so it is difficult to give a complete formal unambiguous definition of the trust [11, 12]. Figure 6.3 shows the categories of trust based on observation, property, cooperation and place of usage. In WSNs, trust plays an essential role in improving reliability as well as cooperation among sensing devices. Table 6.1 lists the various definition of trust frequently used in sensor networks. Trust value is a level (quantification or measure) of belief of one entity towards another entity. Note that in this thesis, trust and reputation are used interchangeably since both estimate the trustworthiness. The decisions made in sensor networks are highly dependent on collected sensitive data; hence, sensor networks must be secure from malicious attacks, including DoS attacks, routing attacks for accurate decision-making. However, the traditional security mechanisms, including cryptographic techniques, are unable to provide security against insider attacks due to resource limitation and deployment nature in a hostile environment. Sensing devices are deployed in the unattended environment for long periods where adversaries can physically capture the sensing device to obtain the private cryptographic keys and other sensitive information stored in the internal storage of the sensor [9]. SNs become legitimate network members by possessing cryptographic keys to

Catagories of Trust

Direct Trust Based on observation Indirect Trust Social Trust Based on property QoS Trust

Based on cooperation

Communication Trust Data Trust

Based on the place of usage

Figure 6.3  Categories of trust.

Behavioural Trust Computational Trust

SE-TEM for WSNs  117 Table 6.1  Various definitions of trust in WSN. Reference/source

Trust definitions

[13]

Trust is the reputation of the entity where reputation is opinion about others.

[14]

Generally, an entity can be said to ‘trust’ a second entity when the first entity makes the assumption that the second entity will behave exactly as the first entity expects.

[15]

Assured reliance on the character, ability, strength, truth of someone or something.

[17]

Trust is a belief [45] that ensures the entity as secure and reliable.

[18–21]

Trust specifies the reliability or trustworthiness of sensor node.

[22]

Trust is an important factor to ascertain reliability of a node.

launch internal attacks. If communicating SNs are captured before the exchange of the cryptographic key, any successive security methods are rendered ineffectual. All the participating nodes should be trustworthy to establish secure communications in sensor networks. Furthermore, cryptographic mechanisms are heavyweight techniques and do not consider the behavior of sensing devices. In order to mitigate insider adversaries, trust-based security solutions have been suggested as an essential lightweight tool for dependability enhancement [22, 23]. Trust-based security solutions are also known as trust models that detect abnormal activities of sensing devices and enhances the trustworthiness [25]. Trust is dynamic, context-dependent, as well as complex concept in WSNs. Trust is reflexive, asymmetric, and transitive [14, 16]. A brief description of certain trust properties as shown by Figure 6.4 are listed below. Reflexive: Every entity (sensing device or node) always trusts itself. Hence, trust is reflexive. Asymmetric: If an entity (node) X trusts entity Y, then it is not necessarily true that entity Y also trusts entity X. Trust of X on Y and trust of Y on X is independent. Hence, trust is asymmetric and unidirectional. Partially transitive: If entity X trusts entity Y and entity Y trusts node Z, then entity X needs not trusts entity Z depending on the behavior of Z, i.e., X may distrust Z. Dynamic: Trust varies with time, depending on the behavior (evidence or experience) of entities. Sometimes, trust varies with the mobility of nodes. Context-dependent: Trust is context-dependent, i.e., trust parameter varies according to context. The context may be different application domain, spatial and temporal context, etc. Subjective and personalized: Trust is being computed based on observation, evidence, and recommendation made available to the SN in a specific situation. Trust models monitors the behavior of the sensing device, estimates the trust value, and then quantified it into highly trusted, trusted, and distrusted. Trust models assure that all the nodes taking part in the communications are trustworthy. The trust (reputation) models proposed in the last few years for distributed systems such as WSNs, multi-agent systems, peer-to-peer networks, ad-hoc networks, have certain key processes in common such as gathering behavioral information, scoring, recommendations, ranking, rewarding,

118  The Smart Cyber Ecosystem for Sustainable Development

Reflexive

Asymmetric

Context Sensitive

Trust

Subjective

Partial Transitivity

Figure 6.4  Properties of trust in WSNs.

and punishing [12, 13]. There are several issues in developing an ideal and proficient trust model for distributed and heterogeneous sensor networks. Some issues are listed as i) what entities (nodes) are interacting and ii) what are the most reliable entities in terms of reputation and confidence etc. have to be taken into consideration when designing such models. WSNs nodes become malicious when they provide false information about other neighbors SNs in terms of their current and past behavior. The key role of trust models and reputation systems is finding the most trustworthy nodes as well as discouraging participation of selfish malicious nodes based on the information provided in order to make good-quality decisions (i.e., distinguish between trustworthy and selfish) in various critical real-time applications such as defense, area monitoring, intrusion detection, and tracking. There are several trust models suggested recently for WSNs to improve the dependability, robustness, and accuracy. We try to identify the characteristic of these trust models in terms of scalability, methodology, complexity, and suitability. The main goals of the trust/reputation system are establishing trust in the wireless communication network that provides several benefits listed below [30–45]: • Trust provides a solution for granting corresponding access control based on the quality of the SNs and their services, which cannot be solved through traditional security mechanisms. • Trust assists routing by providing reliable routing paths that do not contain malicious, selfish, or faulty nodes. • Trust assessment detects erroneous data and allows us to separate (filter out) it from trusted data. The erroneous data may be caused intentionally by attackers or unintentionally due to malfunction of hardware. • Trust assessment identifies selfish and compromised sensing devices and alarm to replace it for proper functioning.

SE-TEM for WSNs  119 • Trust assessment supports energy optimization as well as improve the performance of WSN. • Trust assessment detects various kinds of unexpected behaviors such as an on-off attack, garnished attack, and SN’s failure detection. • Trust assessment improves reliability, decision-making capability, and dependability among SNs, reliable routing paths, key distribution, and data aggregation, lifetime as well as the resilience of a WSN. • Trust assessment discourages participation of selfish node(s) to diminish the damage caused by internal attacks. • Trust assessment estimates the reputation and reliability of SNs for qualitative decision-making. • Trust assessment protects the sensor network from internal adversaries to improve the dependability, robustness, and accuracy of the trust system. Direct trust computation protects the network from rumor spreading but suffers from severe limitations such as i) allowing selfish SNs for a longer time in the system and ii) reputation building time is relatively high/low, respectively. On the contrary, indirect trust (second-hand information) offers several benefits such as stable trust values (over time), quick reputation building as well as swift learning of SNs from each other’s past experience but impose severe burden of i) implementation of reputation exchange protocol, ii) high bandwidth consumption, and iii) transmission energy consumption. To alleviate any internal attacks of WSN, we need to monitor associated misbehavior. As the monitored misbehavior increased, the probability of achieving higher security is increased but needs to check the feasibility with node capabilities. Trust computation based on success ratio requires less processing cost than other schemes where an aging factor is adopted since handling of history require high processing, energy, as well as memory requirement. Depending on the place of usage, trust is classified as behavioral trust and computational trust. Behavioral trust defines trust relations among people and organizations. Computational trust defines trust relation among devices, computers, and networks [14, 29]. Any standard trust model should be designed, having these five components in mind as shown in Figure 6.5. During information gathering process, a SN collects information about other interested nodes by directly observing or indirectly observing. The available neighbor nodes’ information is stored in the routing table of the node. After information gathering, trust evaluation mechanisms such as analytical or fuzzy-based [6] mechanisms are used to compute the reputation value of participating SNs. Figure 6.6 shows the motivation, designing criteria,

Elements of Trust (Reputation) Models

Information Gathering

Scoring and Ranking

Figure 6.5  Elements of trust model.

Entity Selection

Transaction

Reward and Punish

120  The Smart Cyber Ecosystem for Sustainable Development Trust in wireless sensor networks

Motivations

Types (different classification)

Designing Criteria

Security challenges & establishment

Cost

Improving cooperation (dependability)

Convergence rate and execution speed

Detecting internal attacks

Accuracy

Attacks against trust management schemes

Static

On-off

Dynamic

Sybil

Direct

Bad-mouthing

Indirect

Greyhole

Energy

Blackhole

Centralized, Semicentralized or Distributed

Garnished

Proactive and Reactive

Collusion

Delay and traffic overhead Implementation Feasibility and capability Topology

Detecting Malicious (selfish) sensor nodes

Resource consumption (power, storage)

Security challenges due to hostile environment

Scalability

Prediction of future behaviour of sensor nodes

Stability

Insecure, free and unreliable communication

fault tolerance

Reliability

Security problems of TMS

Computational complexity

Mobility Shared and unstable, resource limitation Communication overhead imposed by security mechanism

Adaptability, simplicity and practicality

Periodic

New-comer

High Density Nature and architecture of TMS

Spoofing

Figure 6.6  Motivation, designing criteria, types, and attacks against trust in WSN.

types, and attacks against trust in WSN. TEAs are mainly used to enhance reliability, security, and cooperation of SNs and play a key role to improve system efficiency [3, 4, 6–8, 11, 12]. Trust is a level of assurance or confidence of one entity on another entity. In [11, 15, 19], trust along with its applications are defined in several ways as trust is highly useful in effective decision-making in terms of accepting sensed data and forwarding packets. Trust is useful in terms of providing reliable data paths and corresponding access control. There exist several popular clustering approaches such as LEACH, HEED, EEHC, and EC [2, 5, 14, 16] to group the nodes and elect the cluster head (CH) to improve network scalability and throughput. CH is considered as the most powerful and resourceful node, which is used to detect malicious nodes by estimating the trust degree of ” neighboring SNs.

6.1.3 Major Contribution The proposed approach (SE-TEM) uses a distributed TEA within the cluster and centralized approach with clusters to improve the accuracy of our SE-TEM. Moreover, the proposed method (SE-TEM) efficiently identifies selfish SNs and provides prevention by employing flexible punishment coefficient that can be tuned according to application requirement (ɳ ∈ [1 2]). For simplicity, we have taken parameter ɳ = 1 and ϕ ∈ [1 1.5] to reduce the immense sudden changes in trust values. It will reduce the effect of internal threats. The proposed scheme (SE-TEM) is independent of any specific platform and routing scheme as well as employs a flexible and straightforward (simple) approach for trust estimation. It reduces communication overhead by applying clustering approach along with eliminating the untrustworthy feedbacks.

SE-TEM for WSNs  121 The remaining part of the research paper is organized into various sections, as Section 6.2 discusses related work, and section 6.3 provides some definitions, assumptions, and domain of trust values. Section 6.4 discusses the proposed trust model, and Section 6.5 presents experimental results and their analysis. Finally, Section 6.6 provides a conclusion and future work.

6.2 Related Work There are various existing trust models categorized into centralized, distributed, and clustered trust models. Centralized and distributed trust models are“not suitable for resource constraint (speed and memory) WSN because of ” single point of failure and memory overhead issues, respectively. In this section, we provide a literature review of some existing state of art clustered trust models with their research gap. Bao et al. [6] proposed a “highly scalable cluster-based hierarchical trust management protocol known as HTMP for WSNs to deal with selfish or malicious nodes” effectively. Trust estimation model of HTMP employs a distributed approach along with multi-dimensional trust attributes concept. Authors consider QoS trust, social trust, and probabilistic model to evaluate and validate the performance of HTMP. This scheme is unrealistic because challenging (difficult) to implement at each cluster member (CM) as well as does not deal with false (untrustworthy) recommendations. Zhang [10] proposed a dynamic, hierarchical trust model known as TMA, which employs a decay function and time window concept to assess the dependability of SNs with minimum communication and memory overheads.TMA uses direct and indirect trust along with energy consideration of SNs and a dynamic weighted approach to give more weight-age to recent communication (observation or trust values). Moreover, the author considers “packet drop rate, latency attribute, certificate, and behavior-based trust in trust” estimation. The major limitations of TMA are i) no punishment coefficient, ii) linear trust function, and iii) does not deal with the false (untrustworthy) recommendation. Shaikh et al. [11] proposed a clustered trust model known as GTMS to reduce communication and memory overhead. GTMS uses a robust hybrid trust model to identify and defend various faulty SNs. The author states that GTMS is suitable for large-scale WSN, but it has several limitations such as light and static punishment coefficient during direct trust computation, significant communication overhead due to broadcast-based strategy to collect feedbacks from CMs that makes it unrealistic for different critical applications such as defense and environmental monitoring. Li et al. [12] proposed a novel and efficient trust management approach known as “LDTS: A lightweight and dependable trust system for clustered WSNs.” LDTS reduces communication and memory overheads by canceling the feedbacks between CH and CM, the effect of selfish SNs by using a punishment coefficient. Moreover, LDTS employs a self-adaptive weighted approach for trust aggregation at the CH level. Theoretical and experimental results exhibit the feasibility of LDTS. The static punishment coefficient during trust computation is one of the major drawbacks exists in this trust model that makes it unrealistic. Talbi et al. [13] designed a novel trust management scheme (TMS) known as “Adaptive and dual data-communication trust (ADCT) scheme for clustered WSNs.” ADCT is

122  The Smart Cyber Ecosystem for Sustainable Development realistic trust model that considers data trust and communication trust to effectively cope with malicious (selfish) nodes and untrustworthy recommendations. ADCT uses nonlinear trust function with severity parameter to monitor the behavior of SNs according to the requirements of the application. Experimental results demonstrate the great performance in terms of attack detection, prevention, mitigation of selfish (untrustworthy) nodes, cooperation, and resource efficiency to protect WSN. ADCT does not employ energy trust of SNs and load balancing among Clusters, which should be considered to obtain robust trust values. Moreover, it is not suitable for on-off attack mitigation. Singh et al. [14] designed a realistic trust framework for large-scale WSN is known as “A lightweight trust mechanism and overhead analysis (LWTM) for clustered WSN” to protect it against malicious behavior. LWTM employs priority concepts, trust updating mechanism, reward and penalty, and self-adaptive weighted method to obtain accurate trust values of SNs. Experimental results (using MATLAB) demonstrate that LWTM requires less communication and memory overhead than GTMS and LDTS. The author claims that LWTM can effectively detect and mitigate 35% malicious nodes, which is better than other existing trust models. LWTM also suffers from various limitations such as i) On-off attack and collusion attacks are not discussed, suitable for few attacks, and ii) misbehavior rate and their frequency is not considered. Górski et al. [18] provide a novel approach known as “WSN Cooperative Trust Management Method (WCT2M)” to protect WSN against various cyber threats. WCT2M “uses validity history and deviation history along with direct and indirect trust to obtain a robust and accurate trust value. WCT2M considers numerous attacks to measure the performance of the proposed idea in terms of detection and isolation of selfish SNs in” multi-layer (ML) WSN. WCT2M suffers from Gossip problem, switching mode (when to broadcast. Direct send) problem and various attacks are also not considered.

6.3 Network Topology and Assumptions In WSN model, SNs are classified as CM or CH, as shown in Figure 6.7 (SNs represented by circles are CM and rectangle are CHs). A CM can communicate via single-hop or multi-hop with its CH and CHs forward aggregated data to BS. We assume that nodes are grouped into clusters using well-known clustering algorithms [2, 16, 17], and the CH is selected using schemes mentioned in [10, 13]. We also consider a timing window mechanism used in [12, 13, 20, 21] to count successful and unsuccessful interactions between SNs at and domain of trust values is [0 to 4] which plays a significant role in memory overhead reduction [5, 12].

6.4 Proposed Trust Model The proposed trust model (SE-TEM) works in two levels known as intra-cluster (CM to CM direct and indirect trust) and inter-cluster (CH to CH direct and BS to CH feedback trust). The detailed working of the SE-TEM is discussed in the following subsections. Note that we are dealing only with direct or indirect communication trust at each level.

SE-TEM for WSNs  123

Intra-Cluster Trust Inter-Cluster Trust SN SN

SN SN

SN

SN

SN

SN

CH

CH

CH

SN

Cluster Head

Base station

SN

SN

SN

SN

Sensor Node

SN

SN

Cluster

SN

SN SN

CH

SN

SN SN

SN

Figure 6.7  Clustered WSN topology.

6.4.1 CM to CM (Direct) Trust Evaluation Scheme Trust evaluation at CM when Sx,y (∆t) > 0 & (Ux,y (∆t) > 0 is defined by Equation (6.1) as follows:

Tx,y (∆t) = 4 ×

(

Sx,y (∆t) (Sx,y (∆t) + Ux,y (∆t))

)

1 ϕSx,y(∆t) η Ux,y (∆t)+1

(6.1)



where Δt represent time window to count number the number of successful interactions (Sx,y(∆t)) and unsuccessful interactions (Ux,y (∆t)) between node x and y. The term 1 U x ,y (Dt) + 1 is used to give strict penalty with the increment in non-cooperative interacS x ,y ( Dt) is used to gradually increase (i.e., reward) the trust value with the tions. The term φ increment in successful interactions. Figure 6.8 represents the successful interaction between two SNs interacted directly to obtain direct trust value (First-hand information). The value of the parameter ϕ ∈ [1 1.5] can be set depending on the application requirement. When Sx,y(∆t) > 0 & (Ux,y(∆t) = 0,

124  The Smart Cyber Ecosystem for Sustainable Development Data Packet Transmission

Forwarding

Sensor Node x

Sensor Node y Copy forwarded to Node x

Figure 6.8  Successful interaction between SN x and SN y.

then we set Tx,y (∆t) = 4, i.e., Trusted CM, and when Sx,y(∆t) = 0 & (Ux,y(∆t) > 0, then we set Tx,y(∆t) = 0, i.e., highly untrusted. When ((Sx,y(∆t) = 0 & (Ux,y(∆t) = 0), then Tx,y(∆t) = FTx,y(∆t), i.e., compute “feedback trust (FT) or second-hand information toward node y reported by CH” because of no direct interactions between CMs x and y.

6.4.2 CM to CM Peer Recommendation (Indirect) Trust Estimation (PRx,y(∆t)) In peer recommendation trust evaluation, we only consider direct trusted neighbors to improve the accuracy of SE-TEM as well as to reduce communication overhead since trusted SNs will be less than or equal to the total number of SNs. We define the peer recommendation (indirect communication trust PRx,y(∆t)) of node y maintained at node x at time ∆t by the following equation (6.2):



  PR x ,y c(Dt) =  

∑

 Tx ,j × Tj,y    (6.2) j=1   z 

z

where z ∈ set of directly trusted nodes. Note that directed trust nodes are those SNs whose trust value can be computed directly based on the interactions. Here, we do not consider nodes having Tx,y(∆t) < 2, i. e., eliminate malicious nodes to obtain robust trust value. Final T (absolute) trust score (f x , y (Dt )) is estimated by simply aggregating Equation (6.1) and Equation (6.2) (because it is already proved that “simple averaging performs better than complex averaging” [12, 20]) as defined by Equation (6.3)



(f xT,y (Dt)) =

Tx ,y (Dt) + PR x ,y (Dt) (6.3) 2

Based on the numerical values obtained by Equation (6.3), SN status (S) can be quantized as follows:



 ( 3;4 ) highly trusted node    S(f xT,y ( t)) =     ( 0; θ )        malicious node    ( θ;3)  legitimate node  

SE-TEM for WSNs  125 The value of θ plays a significant role to make SE-TEM suitable for various real-time critical applications because its value can be adjusted (set) at any point time depending on the requirement of applications.

6.4.3 CH-to-CH Direct Trust Estimation The direct trust between CHi and CHj is defined in a similar way as at CM level as follows: TCHi,CHj (∆t) = 4 ×



(

SCHi,CHj (∆t) UCHi,CHj (∆t))

)

1

(6.4)

ϕSCHi,CHj (∆t)

η UCHi,CHj (∆t)+1



6.4.4 BS-to-CH Feedback Trust Calculation In order to obtain CHs trust values, BS periodically sends a request packet to CHs (suppose m) in the same fashion as CH sends to CMs. In response to request packet, CHs forward their direct trust values to the BS. In order to compute FT value, the BS maintains these values into an m*m matrix as follows:



 CH1,1  CH 2,1 B=  ...   CHm ,1

CH1,2

...

CH1,m

CH 2,2

...

CH 2,m

... CHn−1,2

... ...

... CHm ,m

     

In order to compute feedback trust FTBS,CHj (∆t), we have employed the beta distribution function [22]. FBS,CHj (∆t) can be computed using Equation (6.5) as follows:



FTBS,CH j (Dt) = 4 ∗

p +1 (6.5) p+ b+ 2

T where p is positive feedback and b is negative feedback. A global trust value (GCH i ,CH j (Dt)) can be obtained at CHs as follows:



T (GCH i ,CH j (Dt)) =

α ∗ TCHi ,CH j (Dt) + β ∗ FTBS,CH j (Dt) (6.6) 2

where α and β is respective weight-age and depending upon the application requirement, α and β will give more flexibility to select appropriate weight-age for the robust trust model.

126  The Smart Cyber Ecosystem for Sustainable Development

6.5 Result and Analysis This section discussed the outcomes of the SE-TEM model in terms of severity analysis and malicious node detection.

6.5.1 Severity Analysis This section examines the robustness of trust model in terms of severity analysis, attack detection, and mitigation. MATLAB is used to assess the performance of the SE-TEM. Figure 6.9 shows that the proposed trust model can detect various attacks like badmouthing, black hole, garnished attack, and on-off attack by setting the suitable value of ϕ. Moreover, penalty coefficient is directly proportional to the non-cooperative interactions,” i.e., more the unsuccessful interactions, more the strict punishment. The parameter provides the flexibility to punish SNs strictly or usually, whose value can be tuned according to network scenario and application requirements.” Figure 6.10 clearly shows the variations in trust values according to success rate and ϕ and validate that it can catch those nodes whose interaction rate is low or whose behavior is changed frequently (high to low and low to high). We have compared proposed trust model with ADCT [13] to analyze the effectiveness in terms of average trust value because it is better than LDTS [12], GTMS [5] in terms of severity, threat discovery, attack mitigation, and communication overhead. Average trust values are calculated by taking the average of trust values at ϕ = [1 1.5]. Figure 6.11 shows that the SE-TEM gradually increases the trust values with a gradual increment in success rate. We can control this convergence rate of trust values by using suitable ϕ value. The increment in Average trust values of ADCT is always very slow with the increase in success rate that makes it unrealistic for many WSN applications such as monitoring, tracking, and threat detection [23–28]. 6 ϕ=1 ϕ=1.1 ϕ=1.2 ϕ=1.3 ϕ=1.4 ϕ=1.5

5

Trust Value

4

3

2

1

0

0

0.1

0.2

0.3

0.4

0.5 0.6 Success rate

Figure 6.9  Change in trust values with respect to ϕ.

0.7

0.8

0.9

1

SE-TEM for WSNs  127

6

Trust values

5 4 3 2 1 0 12

10

8

6

4

Success Rate

2

0

1

2

5

4

3

6

Range of ϕ values

Figure 6.10  Trust value vs. success rate and ϕ. Comparison of Success rate representation in Proposed work and traditional work 4

Average Trust Value

3.5

ADCT SE-TEM

3 2.5 2 1.5 1 0.5 0

0

0.1

0.2

0.3

0.4 0.5 0.6 Success rate

0.7

0.8

0.9

1

Figure 6.11  Average trust value vs. success rate.

However, ADCT is suitable for critical application like defense applications where robust trust value is required, but SE-TEM is capable of providing trust values “according to application requirements with the help of ϕ”.

6.5.2 Malicious Node Detection In order to determine the robustness of SE-TEM against attack detection, we have injected up to 50% malicious node in WSN consisting of 100 to 500 nodes. Figure 6.12 shows that up to 60% of malicious nodes are easily caught. Note that false positive is the measure in which selfish node provide good reputation about bad nodes and in the similar way false negative can be defined as the measure in which selfish node provide untrustworthy reputation about good nodes.

128  The Smart Cyber Ecosystem for Sustainable Development

% of False positive and False negative

Malicious node detection

50 45

60

False positive

40 35

40

30

20

25 20

0

15

–20 6

False negative

4 No. of clusters 2 1

0

10

20

30

40

% of malicious nodes

50

10 5 0

Figure 6.12  False-negative and false-positive alarms.

6.6 Conclusion and Future Work This research paper discussed a cluster-based realistic trust model to increase security, reliability, and dependability (i.e., cooperation) among SNs. The proposed scheme uses a time window concepts and works in two levels, namely, intra-cluster and inter-cluster levels to compute the trust values. The novelty of our trust model is that it considers only trusted recommendations and reduces communication overhead. Strict punishment and parameter ϕ makes it appropriate for real-time applications. Theoretical analysis and Experimental results exhibit great performance of proposed trust model in terms of its severity analysis, communication overhead, attack detection, and mitigation. In the future, we are planning to add an energy trust module in SE-TEM to make it robust trust model for heterogeneous WSN with minimum communication and memory overheads. We are also willing to add energy trust module along with a flexible weight allocation strategy in the proposed model.”

References 1. Momani, M., Trust models in wireless sensor networks: A survey. International Conference on Network Security and Applications, Springer, Berlin, Heidelberg, 2010. 2. Kumar, D., Aseri, T.C., Patel, R.B., EEHC: Energy efficient heterogeneous clustered scheme for wireless sensor networks. Comput. Commun., 32, 4, 662–667, 2009. 3. Singh, K., Singh, K., Aziz, A., Congestion control in wireless sensor networks by hybrid multi-objective optimization algorithm. Comput. Networks, 138, 90–107, 2018. 4. Crosby, G.V., Pissinou, N., Gadze, J., A framework for trust-based cluster head election in wireless sensor networks. Proc. Second IEEEWorkshop on Dependability and Security in Sensor Networks and Systems, pp. 10–22, 2006. 5. Heinzelman, W.B., Chandrakasan, A.P., Balakrishnan, H., An application-specific protocol architecture for wireless microsensor networks. IEEE Trans. Wireless Commun., 1, 4, 660–670, 2002.

SE-TEM for WSNs  129 6. Bao, F., Chen, I., Chang, M., Cho, J., Hierarchical trust management for wireless sensor networks and its applications to trust-based routing and intrusion detection. IEEE Trans. Netw. Service Manag., 9, 2, 169–183, 2012. 7. Ganeriwal, S. and Srivastava, M.B., Reputation-Based Framework for High Integrity Sensor Networks. Proc. ACM Workshop Security of Ad Hoc and Sensor Networks (SASN ‘04), pp. 66–67, 2004. 8. Yao, Z. and Doh, Y., PLUS: Parameterized and Localized Trust Management Scheme for Sensor Networks Security. Proc. Third IEEE Int’l Conf. Mobile Ad-Hoc and Sensor Systems (MASS ‘06), pp. 437–446, 2006. 9. Boukerche, A., Li, X., EL-Khatib, K., Trust-based security for wireless ad hoc and sensor networks. Comput. Commun., 30, 2413–2427, 2007. 10. Zhang, J., A trust management architecture for hierarchical wireless sensor networks. IEEE Local Computer Network Conference, 2010. 11. Shaikh, R.A., Jameel, H., d’Auriol, B.J., Lee, H., Lee, S., Group-based trust management scheme for clustered wireless sensor networks. IEEE Trans. Parallel Distrib. Syst., 20, 11, 1698–1712, 2009. 12. Li, X., Zhou, F., Du, J., LDTS: A lightweight and dependable trust system for clustered wireless sensor networks. IEEE Trans. Inf. Forensics Secur., 8, 6, 924–935, 2013. 13. Talbi, S., Adaptive and dual data-communication trust scheme for clustered wireless sensor networks. Telecommun. Syst., 65, 4, 605–619, 2017. 14. Singh, M., Sardar, A.R., Majumder, K., Sarkar, S.K., A lightweight trust mechanism and overhead analysis for clustered WSN. IETE J. Res., 63, 3, 297–308, 2017. 17. Younis, O. and Fahmy, S., HEED: A hybrid, energy-efficient, distributed clustering approach for Ad-Hoc sensor networks. IEEE Trans. Mobile Comput., 3, 4, 366–379, 2004. 15. Ng, H.S., Sim, M.L., Tan, C.M., Security Issues of Wireless Sensor Networks in Healthcare Applications. BT Technol. J., 24, 2, 138–144, 2006. 16. Jin, Y., Vural, S., Moessner, K., Tafazolli, R., An energy-efficient clustering solution for wireless sensor networks. IEEE Trans.Wireless Commun., 10, 11, 3973–3983, 2011. 18. Górski, J. and Turower, A., A Method of Trust Management in Wireless Sensor Networks. Int. J. Secur. Privacy Trust Manage., 7, 1–19, 2018. 19. Gautam, A.K. and Kumar, R., A Robust Trust Model for Wireless Sensor Networks. 2018 5th IEEE Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON), 2018. 20. Ishmanov, F., Kim, S., Nam, S., A secure trust establishment scheme for wireless sensor networks. Sensors, 14, 1, 1877–1897, 2014. 21. Ishmanov, F., Kim, S., Nam, S., A robust trust establishment scheme for wireless sensor networks. Sensors, 15, 3, 7040–7061, 2015. 22. Whitby, A., Jøang, A., Indulska, J., Filtering out unfair ratings in bayesian reputation systems, in: The Autonomous Agents and Multi-Agent Systems 2004 (AAMAS2004), New York, 2004. 23. Khan, T., Singh, K., Abdel-Basset, M., Long, H.V., Singh, S.P., Manjul, M., A Novel and Comprehensive Trust Estimation Clustering-Based Approach for Large Scale Wireless Sensor Networks. IEEE Access, 7, 58221–58240, 2019. 24. Aziz, A., Singh, K., Osamy, W., Khedr, A.M., Effective algorithm for optimizing compressive sensing in IoT and periodic monitoring applications. J. Network Comput. Appl., 126, 12–28, 2019. 25. Aziz, A., Singh, K., Osamy, W., Khedr, A., Optimizing Compressive Sensing Matrix using Chicken Swarm OptimizationAlgorithm. IET Wireless Sens. Syst., 9, 5, 306–312, 2019. 26. Aziz, A. and Singh, K., Adaptive compressive sensing based routing algorithm for internet of things and wireless sensor networks. Communication and Computing Systems: Proceedings of the

130  The Smart Cyber Ecosystem for Sustainable Development International Conference on Communication and Computing Systems (ICCCS 2016), Gurgaon, India, vol. 9–11 September 2016, CRC Press, p. 395, 2017. 27. Aziz, A. and Singh, K., Lightweight Security Scheme for Internet of Things. Wireless Pers. Commun., 104, 2, 577–593, 2019. 28. Shi, E. and Perrig, A., Designing Secure Sensor Networks. IEEE Wireless Comm., 11, 6, 38–43, 2004. 29. Gautam, A.K. and Kumar, R., A Robust Trust Model for Wireless Sensor Networks. 2018 5th IEEE Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON), IEEE, 2018. 30. Desai, S.S. and Nene, M.J., MITE: Memory integrity based trust evaluation in Wireless Sensor Networks. 2015 International Conference on Communication Networks (ICCN), IEEE, 2015. 31. Desai, S.S. and Nene, M.J., Node-Level Trust Evaluation in Wireless Sensor Networks. IEEE Trans. Inf. Forensics Secur., 14, 8, 2139–2152, 2019. 32. Zhan, G., Shi, W., Deng, J., Tarf: A trust-aware routing framework for wireless sensor networks. European conference on wireless sensor networks, Springer, Berlin, Heidelberg, pp. 65–80, 2010. 33. Zhan, G., Shi, W., Deng, J., Design and implementation of TARF: A trust-aware routing framework for WSNs. IEEE Trans. Dependable Secure Comput., 9, 2, 184–197, 2011. 34. Thenmozhi, E. and Audithan, S., Trust based cluster and secure routing scheme for wireless sensor network. Second International Conference on Current Trends In Engineering and Technology-ICCTET, vol. 2014, IEEE, pp. 489–494, 2014. 35. He, S. and Zhao, H., Trust and potential field-based routing protocol for wireless sensor networks. 2015 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC), IEEE, pp. 1–5, 2015. 36. Ahmed, A., Bakar, K.A., Channa, M., II, Haseeb, K., Khan, A.W., TERP: A trust and energy aware routing protocol for wireless sensor network. IEEE Sens. J., 15, 12, 6962–6972, 2015. 37. Tajeddine, A., Kayssi, A., Chehab, A., Elhajj, I., Itani, W., CENTERA: a centralized trust-based efficient routing protocol with authentication for wireless sensor networks. Sensors, 15, 2, 3299–3333, 2015. 38. Liu, Y., Dong, M., Ota, K., Liu, A., ActiveTrust: Secure and trustable routing in wireless sensor networks. IEEE Trans. Inf. Forensics Secur., 11, 9, 2013–2027, 2016. 39. Ahmed, A., Bakar, K.A., Channa, M., II, Khan, A.W., A secure routing protocol with trust and energy awareness for wireless sensor network. Mobile Networks Appl., 21, 2, 272–285, 2016. 40. Ahmed, A., Bakar, K.A., Channa, M., II, Khan, A.W., Haseeb, K., Energy-aware and secure routing with trust for disaster response wireless sensor network. Peer-to-Peer Networking Appl., 10, 1, 216–237, 2017. 41. Tang, J., Liu, A., Zhang, J., Xiong, N.N., Zeng, Z., Wang, T., A trust-based secure routing scheme using the traceback approach for energy-harvesting wireless sensor networks. Sensors, 18, 3, 751, 2018. 42. Sun, Z., Zhang, Z., Xiao, C., Qu, G., DS evidence theory based trust ant colony routing in WSN. China Commun., 15, 3, 27–41, 2018. 43. pal Singha, S. and Sharmab, S.C., A Survey on Cluster Based Routing protocols in Wireless Sensor NetworNs. Procedia Comput. Sci., 45, 687–695, 2015. 44. Guo, W. and Zhang, W., A survey on intelligent routing protocols in wireless sensor networks. J. Network Comput. Appl., 38, 185–201, 2014. 45. Kumar, A., Shwe, H.Y., Wong, K.J., Chong, P.H.J., Location-based routing protocols for wireless sensor networks: A survey. Wireless Sens. Netw., 9, 1, 25–72, 2017.

7 Smart Applications of IoT Pradeep Kamboj1*, T. Ratha Jeyalakshmi2, P. Thillai Arasu3, S. Balamurali4 and A. Murugan5 Department of Computer Science and Engineering, North Eastern Regional Institute of Science and Technology, Nirjuli, Itanagar, Arunachal Pradesh, India 2 Department of Computer Applications, Sri Sarada College for Women, Tirunelveli, Tamil Nadu, India 3 College of Natural and Computational Sciences, Wollega University, Nekemte, Ethiopia 4 Department of Computer Applications, Kalasalingam Academy of Research and Education, Anand Nagar, Krishnankoil, Tamil Nadu, India 5 Department of Chemistry, North Eastern Regional Institute of Science and Technology, Nirjuli, Itanagar, Arunachal Pradesh, India 1

Abstract

In recent years, the Internet of Things (IoT) proves to be a powerful technology to analyze the world from a different perspective. Though, there are vast applications of IoT in real life, such as monitoring the manufacturing process, identifying the problems, detecting the transportation, which provide the safety and quality materials to the consumers from producers. The advantages and challenges of IoT technology in the cities, health, agriculture, and industry are discussed in this chapter. Keywords:  Internet of Things (IoT), cites, health, agriculture, industry

7.1 Introduction In today’s world with the global inclination toward smarter living, one cannot ignore the importance of Internet of Things (IoT) devices to achieve the implementation of intelligent infrastructure for providing a better framework to build smart cities, healthcare services, agriculture, industry, and so forth. It is the time for the government agencies and industry leaders to accelerate the implementation of intelligent infrastructure while keeping future augmentation into consideration. Smarter cities mean that they provide better services, ensuring rich living standards, availability of sound healthcare system, ease of governance, and many more aspects but without impinge in citizen’s lives. Creating such cities needs rethinking and replacing the existing infrastructure with a new framework, which is built with a blend of contemporary technologies to form intelligent infrastructure around which future expansion will be designed. To obtain such infrastructure, there is a requirement of lots of intelligent devices, which not only collects the data on *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (131–152) © 2021 Scrivener Publishing LLC

131

132  The Smart Cyber Ecosystem for Sustainable Development Education

Agriculture

Mobility

Security

Home

Smart Cities Industry

Healthcare Civil infrastructure

Retail

Energy

Figure 7.1  Smart infrastructure.

regular intervals but also analyzes it to forecast changes and alters its behavior according to changing data patterns. Intercession in technology and policy applies to various aspects of urban living. Urban planning applications can be categorized into hard and soft domains. Hard domains include environment; transport and logistics; energy-grid; public lighting, natural resources, and water management; healthcare; buildings infrastructure; waste management; and public safety. Soft domains include e-governance; social incorporation; education and culture; and the economy. This wide range of applications leads to diverse uses of the term smart city [1]. Smart infrastructure programs are designed and developed around three main components, which include: (1) connected technologies to fabricate interconnected networks, (2) the smartened infrastructure system, and (3) environmental systems that yield allimportant services [2]. Cell phones and sensors play a vital role as information receivers in building smart infrastructure. Real-time updates regarding infrastructure performance are received by Governing agencies through the mean of enabling technologies. IoT-enabled actuators can also be used to manage infrastructure operations with minimal response time to accelerate the emergency response. Figure 7.1 shows an overview of smart infrastructure [3]. The innermost part consists of community and governing agencies, which exploits enabling technologies (middle layer) to monitor, access, and control infrastructure services and environmental resources (outermost layer). Further intelligent infrastructure is the base source in future development and proves to be beneficial inefficient resource utilization, providing convenience to the citizen, and responding changes with agility. In this chapter, applications of cities, health, agriculture, and industry are examined individually.

7.2 Background 7.2.1 Enabling Technologies for Building Intelligent Infrastructure Smart Infrastructure is developed with a blend of enabling technologies integrated. The enabling technologies capable of developing a smart infrastructure include sensors,

Smart Applications of IoT  133 actuators, data transmission, IoT, big data analytics, blockchain, and so forth. A few of them are described below: • Sensors Sensors are among the major components used to develop smart infrastructures. These devices are used to acquire or quantify the environment’s properties and convert them into an electric signal [4]. Sensors are electronic devices that transmit the data to the network computation nodes and act as an interface between the environment and an intelligent control system. Sensors can be fixed or mobile and may be connected with wired or wireless. The type of sensor used depends upon the application for which it is needed. It changes over signs from improvements into a simple or digital form, with the goal that the crude information about identified boundaries are clear by machines and people [5]. There are a variety of applications for which sensors can be used. Some traditional applications deploy them to measure temperature and pressure; whereas, smart sensors like an aquatic sensor, infrared sensor, proximity sensors, gas sensor, chemical sensor, and many more are deployed in advanced applications. Table 7.1 gives a rundown of certain sorts of sensors and their applications [6]. Wireless sensor systems offer answers for various divisions, for example, industry, smart city [7, 8], smart home [9] smart-energy, associated vehicle [10], smart-agriculture [11], associated building and grounds [12], medicinal services [13], and healthcare and security. • Actuators The function of an actuator is to mechanically respond to the environment based on the data acquired by the sensors and the output signal sent to it by the computational device. Actuators have performance constraints associated with them and act as a balancing entity to achieve the desired task efficiently. The actuators can be used to respond in case of disaster with minimum response time, as in the case of a fire detector, the activation occurs immediately after the detection of fire. They have a wide range of applications to develop smart infrastructure. • Internet of Things The IoT is the connection of common devices including electrical and electronic devices which generally act as a transceiver and become a source of data acquisition using an internet connection. It provides an opportunity to exploit the internet connectivity to perform various tasks including remotely controlling, environment monitoring, data sharing, data visualization, and many more such task can be performed. The IoT can integrate with the latest technology and serves in building a smart infrastructure. There is a wide list of applications where IoT can be embedded to perform numerous tasks and are ranging from mitigating network congestion to smart parking. Figure 7.2 shows the construction of IoT for all applications [14]. Communication technology bears a significant role in a powerful association with IoT systems. There are so many existing principles for communications systems. They are classified as communications-enabled devices, Bluetooth, ZigBee, Z-Wave, passive and active radio frequency identification (RFID) systems,

134  The Smart Cyber Ecosystem for Sustainable Development Table 7.1  Summary of agriculture sensor. Types of sensors

Functions

Applications

Optical

Soil properties are measured using light

Properties of soil like moisture content, presence of organic matter, and type of clay are determined using photodiodes and photodetectors

Mechanical

Probes are used to measure soil compaction or mechanical resistance

Force applied by roots while water absorption is detected by tensiometers and is very useful for irrigation intervention

Electrochemical

The presence of specific ion in the soil is detected by electrodes

Concentrations of ions in plants or soil

Dielectric

Uses capacitance to measure the dielectric permittivity of the soil to detect moisture level in it

Time-domain reflectometry (TDR) to sense soil water content or Frequency domain reflectometry (FDR)

Airflow

The permeability of soil is measured

Properties of soil can be measured

Position

Global positioning system (GPS) exploits the satellite’s capabilities to determines the longitude, latitude, and altitude

The GPS detect the accurate position and is prooved to be very helpful in the agriculture domain

Electronic nose

Measure the volatile organic compounds (VOCs)

The electronic nose typically compares the profile of volatile components from plants/fruits using a connected pattern recognition system

Biosensors

To detect the food-borne pathogens and chemical contaminants

These sensors consist of three parts: a component that recognizes the chemical contaminants and produces a signal, a signal transducer, and a reader device

and low power wide area (LPWA) (examples are LoRa, Sigfox, and NB-IoT) [5, 6, 15, 16]. • Big Data Analytics Big data analytics is the blend of tools, techniques, and technology to transform the heterogeneous and large quantity of data into the form that can be used to derive some information. The purpose of analytics is to find hidden patterns, correlations, and other perceptions in the data to ameliorate strategies, practices, or other operations for the well-being of society. Existing data analytics techniques such as data mining are proved to be inefficient in

Smart Applications of IoT  135

Application Layer Buildings

Smart City

Connected car Smart Grid Smart Parking

Healthcare

Internet Wi-Fi

Network Layer

3G/4G

Broadband

Gateway

Gateway

Sensing Layer WSN

Actuators

RFID

Monitor

Figure 7.2  Diagram of construction of IoT for all applications.

handling a large amount of data and needed better techniques like Hadoop MapReduce, Stratosphere, Apache Spark, NoSQL database management system, and so forth for real-time handling of big data • Blockchain Blockchain technology is a kind of ledger that records the provenance of a digital asset in a distributed manner. Distributed Ledger Technology (DLT) is an alias of blockchain and it enables trust between peer-to-peer entities without the use of credible intermediate parties and hence reduces the cost of electronic transactions. The three main constituents of blockchain are blocks, nodes, and miners. Each chain carries multiple blocks and every block is an amalgamation of three elements: • The data in the block. • A 32-bit whole number is generated randomly during the creation of a block and is called the nonce. • A 256-bit number that is united with a nonce and is called the hash. Hash starts with a large number of zeroes.

Miners use the mining process to create new blocks on the chain. Miners are designed to find the nonce that generates the accepted hash and seems to be very efficient in solving this complex task. The base of blockchain lies in decentralization, where no computer or an organization is allowed to own a chain. Rather, a distributed ledger is maintained via nodes to create a chain. Nodes are the electronic device that keeps the copies of the blockchain and keeps the network functional.

136  The Smart Cyber Ecosystem for Sustainable Development

7.3 Smart City The only reason why people talk about smart cities nowadays is due to the huge acceleration in the field of Information and Communications Technology (ICT) and evolution in hardware and software models. Involvement of ICT in various activities in the city assists in production operations and develops a framework for a smart city. The availability of public services for the inhabitants with the efficient use of resources and reducing the environmental impact is the prime consideration while designing the framework for a smart city. A hazy sketch of various components in a smart city is shown in Figure 7.3 [17]. With the significant rise in the world’s population in the past few decades, the living standards of the people are rising drastically. By the next couple of decades, it is predicted that the urban population will take 70% of the total world population and will consume 75% of total resources and energy. This urban shift could lead to the emission of 80% of greenhouse gases and tends to impact the environment negatively and make the concept of a smart city a necessity. One of the primary goals of a smart city is to mitigate the negative effect on the environment caused due to over urbanization. The components, requirements, and characteristics of a smart city are not as such defined but may vary around the globe. International Organization for Standardization (ISO) provides the standards that are abiding by globally, without compromising safety and quality. Ideally, there is no definition of a smart city that is universally acceptable, we can think of a picture of a smart city that accommodates an enumeration of infrastructure and services that narrates the desire of its inhabitants. The objective behind a smart city is developing institutional, economic, social, and physical infrastructure incrementally to enhance the quality of life of its inhabitants by providing job opportunities, a better healthcare system,

Smart Healthcare Smart Governance and Education

Smart Citizen

Smart Building

Smart City

Smart Mobility

Smart Infrastructure

Smart Energy Smart Technology

Figure 7.3  A broad overview of various components needed in a smart city.

Smart Applications of IoT  137 economic security, quality core infrastructure, and a clean and imperishable environment. The base technology to support smart cities includes sensors, ubiquitous computing, and easy availability of data.

7.3.1 Benefits of a Smart City Some of the salient benefits of a smart city are mentioned below; however, the list is by no means exhaustive in nature. ➢➢ Real-time tracking of sensor and surveillance system lead to lesser response time in mitigating the public safety threats. ➢➢ Defines a peer-to-peer model for sharing economy and exchanging the product and services. ➢➢ Provides a rich environment to the potential group of citizens to work altogether on a common interest. ➢➢ Collaborative development of decision-making to form a new era of digital democracy. ➢➢ Enhance the ability of decision-making by data-driven technologies. ➢➢ Smart management of transportation infrastructure to alleviate congestion. ➢➢ Superior healthcare infrastructure gives rise to accurate diagnostics and treatment facilities. ➢➢ Real-time tracking of energy usage helps in managing the use of exhaustive resources. ➢➢ Sensor-driven waste management systems assist in maintaining the balance in the environment. ➢➢ Real-time mapping of crime patterns helps in anticipating crime ahead of the actual occurrence of the incident.

7.3.2 Smart City Ecosystem Smart city ecosystem is the amalgamation of technology, people, organization, and governance to produce the desired outcomes. The backbone of a smart city is the ICT infrastructure which is integrated with numerous components to generate the desired outcomes keeping in mind the various attributes of a smart city and work around different themes. The various components of a smart city include smart infrastructure, smart buildings, smart transportation, smart energy, smart technology, smart healthcare, smart governance, smart education, and many more to list. The attributes or outcomes of a smart city comprise sustainability, quality of life, urbanization, health and wellness, public safety, and smartness. The sustainability of smart cities is associated with the viability of components in parallel with climate change and the ability to handle the issues related to energy, pollution, waste management, and health. Quality of life refers to the social and emotional quotient of the citizens. Urbanization encompasses technology, economics, infrastructure, and governance. The smartness of a smart city is related to the improvement in environmental standards, economic, and social status of its inhabitants. The four themes of a smart city include society, economy, environment, and governance. The society theme is associated with the visualization of a city from its inhabitant’s point of view. The economic themes of

138  The Smart Cyber Ecosystem for Sustainable Development SERVICES Parking Lighting Traffic

IES

CIT

Streetlight Water Mgmt Microgrids

Ridesharing Airports Private Commuting Districts Micro Services Bike Sharing Office Parks

NS

S

ITIE

IL UT

IO AT

R CO

R PO

OUTCOMES

IES

NIT

U MM

CO

NS

IZE

CIT

QUALITY OF LIFE GOVERNMENT EFFICIENCY HEALTH AND WELLNESS ECONOMIC DEVELOPMENT SUSTAINABILITY RESILIENCE PUBLIC SAFETY MOBILITY

Innovation Enablement and Acceleration Community Engagement Grovernance, Management and Operations Policies, Processes, Public-Private Partner and Programs Data “Marketplace” and Analytics Connectivity, Accessibility and Security Smart City Technology Infrastructure

Figure 7.4  A smart city ecosystem.

smart cities make strides in the job opportunity and economic growth of their citizens. The environment theme is associated with the sustainability of a smart city for future generations. The governance theme encloses the robustness of a city to manage policies. Figure 7.4 shows how an ICT infrastructure is exploited to realize the theme of a smart city through various components to provide services and with specified outcomes [18].

7.3.3 Challenges in Smart Cities As the cities are growing in terms of smartness, their challenges need to be taken and addressed so that all the components work at the same pace. • Existing job profile and employment: Adoption of smart technologies in smart cities leads to disruption in the existing work environment and demands the skill up-gradation of employees on large scale. Even though, the use of smart technologies may result in the reduction of manpower as most of the work is automated and give rise to a furor among job aspirants. A research was conducted by researchers at Oxford University to study the impact of automation on existing jobs. They took a sample of 700 different jobs and found that 47% of the existing jobs will become obsolete in the next 10 to 20 years. However, it generates new opportunities in new technological areas, but it remains one of the major challenges to deal with. • Social coherence and harmony: Maintaining social coherence is beyond the scope of any government on its own and needs collective efforts from citizens to achieve the same. One side of smart technologies in a smart city gives rise to new ways to connect people, the other but the darker one is that the benefits of a smart city are not realized evenly by all the segments of the society due to the following reasons: ➢➢ Some groups of people may have less attraction toward new technologies and they do not want to adopt them. ➢➢ The perception of people to see the technology as a threat to their health may also appear as a big hurdle in social coherence.

Smart Applications of IoT  139 ➢➢ Smart solutions may be used by the people to organize themselves and restrict the entry for others can be a major threat to social coherence. • Security and privacy: Security and privacy always remain a concern with the advancement in technology. In a smart city, a large amount of data is generated by the IoT devices that can be misused by hackers or cyber-terrorists. People in a smart city always remain connected to the internet through the devices they have in their possession or through some other ways, which give chance to cyber offenders to make a privacy breach and track or observe their activities to commit frauds. The digitization also creates new opportunities for criminals to do the crime. In the digital world with scalable technologies, committing a crime will no longer be limited to physical boundaries. • Resilience: The term resilience here means the capability to determine the vulnerability in the system against unexpected disruptions and the ability to adjust to the changing conditions. The critical infrastructure of a smart city should be made resilient to handle the disruption in the crucial services due to compromised and malicious components. Further, a resilient smart solution identifies the crucial services and ensures the smooth running of these even in case of disruption. A resilient smart infrastructure also minimizes the impact of undesired incidents and natural disasters. The most pertinent example of an unforeseen situation to be quoted here is COVID-19, which makes the whole world comes to a halt and badly impact the social and economic tendency of an individual and country as a whole. To restore to the earlier state of life is a big challenge in a situation like this, even for the most advanced smart cities. The digital infrastructure of a smart city should be of the capability to handle issues like this steadily.

7.4 Smart Healthcare Health is among the top priorities of any nation. A nation cannot be socially and economically stable without healthy citizens. Health leads to the productivity of a person and contributes hugely to building a strong nation. The healthcare system evolves exceptionally with ages and still contains scope for improvement. Introducing smartness in a healthcare system is one such improvement. A smart healthcare facility is an amalgamation of the traditional healthcare system with ICT and advanced medical equipment. A smart healthcare system works in two dimensions; firstly, it mitigates the need for emergency health services by the timely diagnosis of the ailment and secondly, during the treatment it helps in generating alerts to the doctor well before the patient collapse by constantly monitor the condition of a patient using data acquisition and analytical tools, which again contributes in minimizing the need for emergency healthcare services. Figure 7.5 displays the architecture of a smart hospital system [19]. Here, it is mainly categorized into three layers. The top layer consists of an RFID-rich wireless sensor network, termed as Hybrid Sensor Network which senses the data and routes to the middle layer. The middle layer is a gateway, which enables the communication between sensors deployed on the body of the patient and the remote users. Some monitoring application also runs on the gateway to continuously monitor the incoming traffic and stores it in the database. The patient’s information stored in the

140  The Smart Cyber Ecosystem for Sustainable Development

Blood Pressure

Cloud Server

ECG

Pluse Oximetry

Internet EMG

Families

Gateway

Humidity Temperature

hospital room 1 hospital room 2

Doctors

Figure 7.5  Architecture of smart hospital system.

database is easily accessible to the local doctors via the user interface and also to the remote users with some specific privilege. To illustrate how intelligent digital infrastructure helps in minimizing the need for emergency services, consider the following scenarios: • A person X lives alone in a house that is equipped with smart electric and cooking gas meters. These meters monitor energy usage in real-time by collecting data regarding electricity and gas consumption. The data collected by the meter over some time is given as an input to an analytical tool to learn the energy usage pattern, as a result, it gives an alert to the family in case of any deviation from the normal behavior. Suppose, if the energy usage suddenly shows a spike in the night hours then an alert may be generated to indicate a sleep disorder. Alternatively, if the energy usage grows with time, which means the person X spends most of his time staying at home and shows the sign of depression, accordingly an alert may be generated. Consider another example where a person X is a diabetic patient. For a diabetic person, hypoglycemia is among the greatest risk where he may fall unconscious or even lead to coma. Person X is equipped with a smart wearable medical device to monitor his body temperature and position and let him live in a house that is fitted with pressure and motion sensors. These smart devices constantly monitor his movement and may generate an alert if some change in his gait is found or he falls.

7.4.1 Smart Healthcare Applications The advancement in technology and increasing reliability in IoT devices enables the researchers to discover more and more number of healthcare applications. Though the list of smart healthcare applications is too long, some of them are listed in this section.

Smart Applications of IoT  141 • Telemedicine: Telemedicine makes use of the internet to provide clinical consultation through video conferencing in remote areas where physical infrastructure to provide healthcare services is not installed. It removes the distance barriers to provide clinical help to the people in need. • Automated remote patient monitoring: IoT in healthcare gives rise to a new area of networking known as Wireless Body Area Networks (WBANs). WBAN is a connection of various medical sensors deployed on the patient’s body. These sensing devices take the real-time measurement of crucial indicators, like heart rate or glucose level, and transmit the data to the centralized control device. This control device may send the alert messages, as and when required, to the doctor or can give directions to the actuator for automated treatment. As per the study, around 7.1 million patients around the world are currently using the remote patient monitoring system and this figure may grow up to 50.2 million by 2021 [20]. • Lifestyle wearables: Today, smart wearable devices are readily available that keep track of our daily fitness activities and are capable of generating an alert in case of abnormal body response and can even send the current location of the person using the GPS installed in it. Smartwatches, smart gloves, and smart shoes are some examples of such devices. Some devices can be worn in an ear to track the alertness of a person and proves to be very useful while driving a vehicle. • Infectious disease surveillance: Human ability to detect and monitor manually the infectious disease that threatens the health is not enough and there arises a need to make use of technology to contain the spread of such kind of disease. One of the examples of such a disease surveillance system is the Aarogya Setu Mobile App, developed by the Department of Health, Government of India, to contain the spread of COVID-19. This mobile application exploits the Bluetooth and GPS technology to detect the COVID-19 positive patient and alerts the nearby user who comes in contact with the patient and help in containing the spread of the pandemic. • Real-time air quality monitoring: The way IoT devices are transforming human life is commendable and one such example is the use of IoT sensors in monitoring the air quality in real-time. The metro cities in India are equipped with these sensors installed in a different part of the city to constantly monitor the air quality and provide the base to the local administration in taking necessary action to improve the air quality.

7.4.2 Challenges in Healthcare Medical science and the healthcare system have seen a revolutionary transformation in the past few decades. The advancement in the various aspects of the healthcare system opens new challenges for healthcare providers to sustain in an ever-changing environment. Some of the major challenges in the implementation of healthcare services are discussed below: • Resilience: Patient diagnosis and treatment are always time-critical and in today’s world of smart healthcare; it becomes a challenge to have all smart devices installed to function properly and produce the results at right time.

142  The Smart Cyber Ecosystem for Sustainable Development The network of IoT devices of smart healthcare must comply with the recent technologies, which are used for communication, data sensing, data uploading, patient’s electronic record maintenance, video distribution, and so on. Failure to any of these may lead to impairing the whole smart healthcare. There are several ways to make the whole network more resilient, such as by introducing device level redundancy, power redundancy, and hot-swappable power supply units. • Security: Smart healthcare needs two-way security; firstly, to secure the whole network from internal and external threats, and secondly, to comply with the privacy policies and provide total patient confidentiality. This can be achieved by introducing security elements in the network, such as Unified Threat Management System (UTM), VLANs, and Advanced Traffic Management. • High performance: There are some special network requirements while executing smart healthcare services. The digitization of patient records becomes the priority of healthcare providers and the availability of such records, as and when required is only possible with the high-speed network and tremendous storage capacity in the cloud. The variety of patient records, such as admission information, Electronic Medical Records (EMR), Magnetic Resonance Imaging (MRI), Computing Tomography (CT) scan, ultrasound, X-rays imaging, and many more such records, need high bandwidth required to send the required information without delay. The faster the transmission, the quicker the diagnosis helps in early detection and cure of the patient’s disease. • Flexibility and Scalability: With the rapid growth in technology and increasing demand for smart healthcare services, the healthcare system need to adapt to the changes at the same pace. The whole healthcare system is required to be organized in a way that it becomes future proof. The healthcare network should be flexible enough to accommodate the changes in technology and also can sustain the increasing number of patients and medical staff.

7.5 Smart Agriculture Monitoring is a key factor for taking the right and more precise choices, raising profitability, and nature of the cultivation. Agricultural monitoring should be possible through wired and remote sensors. It has been utilized as far as correspondence, permitting far off admittance to the internet and actualizing sharp calculations for meta-preparing of the data information to improve checking. Yield observing can gather information about the soil conditions, nutrient cycles, disease dissemination, and controlling action by humans, precipitation, and sun-powered radiation. Based on the data, monitoring is utilized to improve the nature of the yield production, limit the chance, and boost benefits. Water level and quality, temperature levels, saltiness, pH level, mugginess, and daylight are some of the factors that affect the development of fishes. The monitoring of these parameters of water helps in more yield of fish culture. Forest assumes a significant job in water and carbon cycles. Obtaining data from the forest, controlling forest fire, photosynthesis, carbon, and water checking have been valuable for controlling the infections and improving the yields of forest products [6].

Smart Applications of IoT  143

Figure 7.6  A modern IoT livestock paradigm.

Observing of domesticated animals relies upon the kinds of creatures viable. For instance, the health condition of dairy animals can be determined by observing the pH level of their milk. The farmers can easily track and question the area of their animals by labeling every single animal with an RFID device, in this way forestalling animal burglary. Wireless sensors have been utilized frequently in animal tracking and social examination. The exhibitions of the gear utilized in animal use have been observed [21]. How an RFID can be utilized in tacking, start from the manufacturing process, handling, stockpiling, selling to consumers [22]. It gives the capacity to gather, store, and analyze data over a large distance at high speed. Figure 7.6 demonstrates how various sensors, which are deployed in the field and on dairy animals, provide you the way to monitor the surroundings of the livestock and consequently helps in better yield and production [23].

7.5.1 Environment Agriculture Controlling Plants are developing wherever whenever giving appropriate natural conditions. This cycle is called greenhouse technology. A horticulture data cloud and an equipment mix of IoT and RFID give the data about precise in anticipating the harvest water needs and furnished cultivators with proficient signs about appropriate time for irrigation process [24]. In Figure 7.7, it is shown how various sensors deployed in the greenhouse transmit the data via cloud servers to the hand-held devices and helps in taking corrective measures accordingly [23].

7.5.2 Advantages The fundamental benefits of IoT in agribusiness are briefly described below. • IoT devices are capable of sensing and generating a large amount of data, which is the primary requirement to monitor the surroundings of community horticulture in urban and rural regions.

144  The Smart Cyber Ecosystem for Sustainable Development

Figure 7.7  An example of cloud IoT solution for greenhouse plantation.

• Automatic water system frameworks that work as indicated by temperature, mugginess, and soil dampness. All these parameters can be gathered at regular intervals through different sensors [25]. • Automatic collection of environmental parameters such as temperature and moistness through sensor systems for additional processing and examination. • Decision emotionally supportive networks that investigate a lot of information to improve operational efficiency and profitability [26, 27]. • The utilization of IoT in farming will assist with setting aside time and cash in reviewing huge fields contrasted with workforce genuinely examining the field either through the utilization of vehicles or walking. • IoT devices are very productive in identifying the bugs infected crop areas and limit the use of pesticides, which, in turn, helps to reduce cost and wastage. • The use of IoT in farming will assist by focusing on the data, and the administration can be done using mobile applications. Likewise, government organizations and administrative standards concerning various homestead produce can be made promptly accessible. Moreover, buyers who are keen on natural products and new products can without much stretch find farmers or be alarmed when new products are accessible. • IoT will empower ongoing checking of homestead resources and machinery against burglary, the substitution of parts, and for opportune routine maintenance [6].

7.5.3 Challenges Most of the farmers from the village are not educated and are poor. They do not know the uses of IoT. The setup and running cost of IoT applications with a computer is very high and

Smart Applications of IoT  145 most of the farmers are incapable to purchase them. Uneducated farmers are unable to do business-related IoT technologies [6]. Farmers cannot solve several technical issues and are not able to update the IoT software. The paucity of acceptable security mechanisms may result in data loss, security and privacy breach, and hindrance in accessing raw data related to various on-field parameters. This may compromise the advantage of deploying advanced IoT devices in the field. So, there arises a need to address several IoT security-related issues, which may occur at various levels of the IoT ecosystem. In agriculture, the IoT devices are prone to get damaged due to various reasons, such as theft, and attacks by animals [28, 29]. The safety of the costly IoT devices deployed in the field of agriculture should be of utmost importance to protect them from climate change [6].

7.6 Smart Industries IoT technology is utilized to assist with directing industrial updating and change to accomplish sharp creation and high worth included manufacturing processes economically and efficiently. Some IoT applications in industries are described below [30]. • Healthcare Service Industry In healthcare systems, IoT devices can be used to monitor patients, places, and medicine [31]. By using IoT technology, it becomes possible to collect, supervise, and share the healthcare-related information efficiently. For instance, a patient’s pulse rate can be captured by a sensor every once in a while and afterward may be transmitted to the healthcare’s office. By utilizing the individualized computing tools (PC, mobile phone, tablet, and so forth.) and uninterrupted internet connectivity, the IoT-based social insurance administrations can be portable and customized [32]. Android mobile-based IoT applications may be monitored in-home healthcare (IHH) services [33]. • Mining Production Mine security is a major worry for most countries because of the operational conditions inside the mines. IoT devices can be deployed to detect mine calamity signals to make prompt communication, disaster forecasting, and safety improvements to prevent and decrease accidents in the mining process. By utilizing RFID and different wireless communication technologies, such as Wi-Fi, correspondence between on-ground and underground becomes viable. Mining organizations can analyze the data provided by different sensors to detect the vulnerable areas and determine the extent to which the excavation can be done without compromising safety. Chemical sensor-based IoT can be deployed on mineworkers’ body for timely detection and diagnosis of disease, as they work in a risky situation. More exploration is required concerning the security attributes of IoT devices utilized in mining production [34]. • Transportation and Logistics Smart transportation, which is more commonly termed as Intelligent Transport System (ITS), makes use of various IoT components, such as

146  The Smart Cyber Ecosystem for Sustainable Development

•



•

sensors, actuators, data analytical tools, data communication methods to enhance safety, leisure, availability, and, on the other hand, reduces the traveling time of the commuters. Smart transportation includes vehicles that are equipped with sensors, GPS, and communication devices and help it to communicate with other intelligent vehicles or even to the roadside intelligent units to pass on emergency and safety messages. In a smart transportation system, vehicles make use of sensors to avoid accidents. In toll collection plazas, sensors which are deployed there, automatically detect the toll charges by sensing the radio frequency identification (RFID) mounted on the vehicles and help in mitigating the congestion on the highways [35]. Advanced IoT applications have been used for automobile services through powerful sensing, networking, communication, and edge computing capabilities. Recently, a renowned car manufacturing company, BMW, developed an intelligent informatics system (iDrive system) to screen the earth, and the street condition to give driving bearings [36]. A smart transportation system also allows its users to select an efficient route to reach the destination and also assists its users to select the best possible transportation option from the available options with minimum cost. In a smart transportation system, airports are also equipped with sensors to automatically detect the identity of the passengers with smart passports in their possession. Smart passports have RFID installed in it. Energy Management and Smart Technology Energy can be defined as an object property to do work and it can neither be created nor destroyed, but it can only be converted from one form to another. Few examples of energy sources include gas, fossil fuels, solar, electricity, and battery. In recent times, people started thinking of different but important aspects of energy, that is, how to make energy consumptions so that it has minimum implications on the environment. This can be achieved by using green energy sources like solar and wind. The use of sustainable energy sources has also attracted the attention of people and administration, to make the energy sources remain available for generations ahead. Technology and artificial intelligence are the key facets in the realization of a smart city. Technology makes the doctors operate their patients even from distant places with the help of a smart robotic arm. IoT technology is used to make smart grids and smart buildings. Another example of technology is the sustainable transport system, which is used to transport a huge number of people to their destinations and proves to help maintain a green environment by reducing road traffic. Smart Infrastructure and Buildings The traditional infrastructure of a city includes roads, buildings, and any such physical component that helps in the functioning of the city and its inhabitants to operate. Smart infrastructure on the other side exploits the digital, electrical, and physical infrastructure. The range of smart infrastructures comprises of waste management system, traffic management system, highway network, rail network, city security and surveillance system, smart electricity distribution system, smart buildings, offices, disaster management system, and so forth.

Smart Applications of IoT  147 Computer via Ethernet Smart Building Cloud Server

Smart Building with Sensors

Mobile Device Computer via WAP

Figure 7.8  Smart building architecture.

A typical architecture of a smart building is shown in Figure 7.8, where the building is equipped with sensors, actuators, and Wi-Fi routers. Various sensors installed in different parts of the building sense the surroundings and transmit the data continuously to the cloud servers. The data analytical tools exploit the data from the server and help in autonomous decision-making [37]. • Food Supply Chain (FSC): IoT technology can be used to monitor the food product, starting from manufacturing till delivery, to guarantee the quality and safety of the product (Figure 7.9) [23]. An average IoT solution for FSC (the supposed Food-IoT) contains three sections: a) on-field devices, for example, RFID readers, sensors, and so on; b) supporting system, for example, databases, servers, and numerous sorts of terminals associated by conveyed PC systems, and so on; and c) the correspondence infrastructures, for example, WLAN, cellular, satellite, power line, Ethernet, and so on [38]. RFID is one of the IoT innovations which are used frequently in FSC. Technological advancement, joined with the expanding strength and development of a few innovations met in IoT, has allowed specialists to create total frameworks, which consolidate detecting modules and programming foundations [23].

7.6.1 Advantages The usage of IoT sensing information empowers the getting of constant data of the physical condition and machine status. By monitoring the manufacturing plant, using IoT devices, it becomes easy to improve the condition of machines and helps to maintain the quality of the product. Besides, the IIoT suite permits a boundless association with the cloud system contrasted and hard-wired system through programmable rationale regulators. The dynamic

148  The Smart Cyber Ecosystem for Sustainable Development Food Processing & Manufacturing Raw Materials

Packaging Consumer

Food Supply Chain Transportation & Logistics

Market & Retail

Figure 7.9  Food supply chain.

of the business rationale and detecting information extraction from the sensing layer are planned with a decentralized methodology. It permits rapid information assortment in the IIoT suite and moves the information to the cloud stage for additional preparation [39].

7.6.2 Challenges The creation of industry application-related IoT programming is exceptionally troublesome based on execution and cost impediments. At the point when a more number of physical objects are associated with the IoT, scalability at various levels of implementation becomes one of the major issues to be addressed [40]. A lot of information transmission over the network simultaneously can likewise cause visit deferral, struggle, and correspondence issues. The IoT network is an extremely convoluted heterogeneous system. The extent of sensing from the IoT suite may drastically affect the data volume needed to be sync in the cloud storage. Hence, the speed of data processing from large data must be as quick as conceivable to meet the quick information stream [39]. Amazing assistance disclosure strategies and item naming administrations should be created to spread the IoT innovation [41]. Dissecting or mining a huge amount of data from IoT applications to determine important information requires a strong foundation of big data analytics. Moreover, incorporating IoT devices with outside assets, for example, existing programming frameworks and Web administrations, needs the involvement of different middleware solutions since applications change a ton by industries. Developing functional applications to join IoT-related information together with conventional information can a tedious task for industries. Explicit issues other than mentioned above are interoperability issues, security and protection issues, and radio access level issues [42, 43]. It ought to be seen that a few issues, for example, the meaning of privacy and legal interpretation are as yet dubious and are not characterized in IoT. Although the current system security innovations give a premise to protection and security in IoT, more work despite everything should be finished [44]. A dependable security insurance instrument for IoT should be explored from the accompanying angles: 1) the definition of security and

Smart Applications of IoT  149 protection from the perspective of social, legitimate, and culture; 2) trust and notoriety component; 3) correspondence security, for example, start to finish encryption; 4) security of communication and user data, and 5) security on administrations and application.

7.7 Future Research Directions In the recent past, we saw thrive in technologies to develop smart infrastructure and we are heading to a future where almost every device will be smart. The involvement of a variety of IoT devices on heterogeneous platforms makes the whole smart infrastructure vulnerable to serious issues, such as virus attacks, data theft, unauthorized user access, eavesdropping, issues related to data integrity, and so forth. These are the issues related to security and privacy which need to be addressed in future research. Further, to handle the massive data generated by IoT devices and to derive the information out of that data, we need to advance research in data analytics.

7.8 Conclusions IoT-enabled technologies are the building blocks of smart infrastructure. A revolution in IoT technologies and its use in smart urbanization have transformed the lives of the people living in it. A new era of urban living has started, where the tendency is to enhance the comfort, safety, and accessibility by providing smart city services to its inhabitants. The main focus of this chapter is to explore IoT-enabled technologies that are used to build smart infrastructures. Further, a detailed discussion of the advantages and challenges in smart cities, healthcare, agriculture, and industries has been done.

References 1. Neirotti, P., De Marco, A., Cagliano, A.C., Mangano, G., Scorrano., F., Current trends in smart city initiatives: Some stylised facts. Cities, 38, 25, 2014. https://doi.org/10.1016/j.cities.2013.12.010. 2. Trindade, E.P., Hinnig, M.P.F., da Costa, E.M., Marques, J.S., Bastos, R.C., Yigitcanlar, T., Sustainable development of smart cities: a systematic review of the literature. J. Open Innov., 3, 1, 2017. https://doi.org/10.1186/s40852-017-0063-2. 3. Alavi, A.H., Jiao, P., Buttlar, W.G., Lajnef, N., Internet of things enabled smart cities: state-ofthe-art and future trends. Measurement, 129, 589, 2018. 4. Hancke, G.P., De Carvalho e Silva, B., Hancke, G.P. Jr., The role of advanced sensing in smart cities. Sensors, 13, 1, 393, 2013. https://doi.org/10.3390/s130100393. 5. Bhuvaneswari, V. and Porkodi, R., The Internet of Things (IoT) Applications and Communication Enabling Technology Standards: An Overview. IEEE-2014 International Conference on Intelligent Computing Application, pp. 324, 2014. 6. Elijah, O., Rahman, T.A., Orikumhi, I., Leow, C.Y., Hindia, M.H.D.N., An Overview of Internet of Things (IoT) and Data Analytics in Agriculture: Benefits and Challenges. IEEE Internet Things J., 5, 5, 3758, 2018. 7. Zanella, A., Bui, N., Castellani, A., Vangelista, L., Zorzi, M., Internet of Things for smart cities. IEEE Internet Things J., 1, 1, 22, 2014.

150  The Smart Cyber Ecosystem for Sustainable Development 8. Sanchez, L., Munoz., L., Galache., A.A., Sotres, P., Santana, J., Gutierrez, V., Ramdhany, R., Gluhak, A., Krco, S., Theodoridis, E., Pfisterer., D., Smart Santander: IoT experimentation over a smart city testbed. Comput. Netw., 61, 217, 2014. https://doi.org/10.1016/j.bjp.2013.12.020. 9. Park, E., Cho, Y., Han, J., Kwon, S.J., Comprehensive approaches to user acceptance of Internet of Things in a smart home environment. IEEE Internet Things J., 4, 6, 2342, 2017. 10. Husni, E., Hertantyo, G.B., Wicaksono, D.W., Hasibuan, F.C., Applied Internet of Things (IoT): Car monitoring system using IBM BlueMix. Proc. Int. Seminar Intell. Technol. Appl. (ISITIA), 417, 2016. 11. Brewster, C., Roussaki, I., Kalatzis, N., Doolin, K., Ellis, K., IoT in agriculture: Designing a Europe-wide large-scale pilot. IEEE Commun. Mag., 55, 9, 26, 2017. 12. Sastra, N.P. and Wiharta, D.M., Environmental monitoring as an IoT application in building smart campus of University Udayana. Proc. Int. Conf. Smart Green Technol. Elect. Inf. Syst. (ICSGTEIS), vol. 85, 2016. 13. Islam, S.M.R., Kwak, D., Kabir, M.H., Hossain, M., Kwak, K.S., The Internet of Things for healthcare: A comprehensive survey. IEEE Access, 3, 678, 2015. 14. Atlam, H.F. and Wills, G.B., Chapter Three-Intersections between IoT and distributed ledger. Adv. Comput., 115, 73, 2019. https://doi.org/10.1016/bs.adcom.2018.12.001. 15. Beyene, Y.D., Jantti, R., Tirkkonen, O., Ruttik, K., Iraji, S., Larmo, A., Tirrronen, O., Torsner, J., NB-IoT technology overview and experience from cloud-RAN implementation. IEEE Wireless Commun., 24, 3, 26, 2017. 16. Raza, U., Kulkarni, P., Sooriyabandara, M., Low power wide area networks: An overview. IEEE Commun. Surv. Tuts., 19, 2, 855, 2017. 17. Samee, S., Khan, S., Vengatesan, K., Bhaskar, M.S., Snajeevikumar, P., Jaspreetkunar, P., Smart city automatic garbage collecting system for a better tomorrow. Int. J. Pure Appl. Math., 114, 9, 455, 2017. 18. https://strategyofthings.io/smart-city-ecosystem, 2018. 19. Chuang, Y.-H., Lo, N.-W., Yang, C.-Y., Tang, S.-W., A lightweight continuous authentication protocol for the internet of things. Sensors, 18, 1104, 2018. 20. Mack, H., Remote patient monitoring market grew by 44 percent in 2016, report says, https:// www.mobihealthnews.com/content/remote-patient-monitoring-market-grew-44-percent2016-report-says, 2017. 21. Jeong, H. and Yoe, H., Study on the sensor MAC protocol for environmental monitoring of livestock farm. Proceedings of international conference on wireless networks (ICWN), The steering committee of the world congress in computer science, computer engineering and applied computing (WorldComp), p. 1, 2012. 22. Gan, W., Zhu, Y., Zhang, T., On RFID application in the tracking and tracing system of agricultural product logistics. International Conference on Computer and Computing Technologies in Agriculture, Springer, Nanchang, China, pp. 400–407, 2010. 23. Tzounis, A., Katsoulas, N., Bartzanas, T., Kittas, C., Review-Internet of Things in agriculture, recent advances and future challenges. Biosyst. Engin., 164, 31, 2017. 24. Tongke, F., Smart agriculture based on cloud computing and IoT. J. Convergence Inf. Technol., 8, 210, 2013. https://doi.org/10.4156/jcit/vol8.iwwue2.26. 25. Kodali, R.K., Jain, V., Karagwal, S., IoT based smart greenhouse. Proc. IEEE Region 10 Humanitarian Technol. Conf. (R10-HTC), pp. 1–6, 2016. 26. Gómez-Chabla, R., Real-Avilés, K., Morán, C., Grijalva, P., Recalde, T., IoT Applications in Agriculture: A Systematic Literature Review, in: ICT for Agriculture and Environment. Second International Conference, CITAMA 2019, Guayaquil, Ecuador, January 22-25, 2019, Proceedings, Advances in Intelligent Systems and Computing 901, R. Valencia-García, G. Alcaraz-Mármol, J. del Cioppo-Morstadt, N. Vera-Lucio, M. Bucaram-Leverone (Eds.), pp. 68–76, Springer, Switzerland, AG, 2019.

Smart Applications of IoT  151 27. Nandyala, C.S. and Kim, H.K., Green IoT agriculture and healthcare application. Int. J. Smart Home, 10, 289, 2016. 28. Cambra, C., Sendra, S., Lloret, J., Garcia, L., An IoT service-oriented system for agriculture monitoring. IEEE International Conference on Communications, pp. 1–6, 2017. 29. Varga, P., Plosz, S., Soos, G., Hegedus, C., Security threats and issues in automation IoT. Proc. IEEE 13th Int. Workshop Factory Commun. Syst. (WFCS), pp. 1–6, 2017. 30. Xu, L.D., He, W., Li., S., Internet of Things in Industries: A Survey. IEEE Trans. Ind. Inf., 10, 2233, 2014. 31. Duan, J., Gao, D., Yang, D., Foh, C.H., Chen, H.H., An energy aware trust derivation scheme with game theoretic approach in wireless sensor networks for IoT applications. IEEE Internet Things J., 1, 1, 58, 2014. 32. Alemdar, H. and Ersoy, C., Wireless sensor networks for healthcare: A survey. Comput. Netw., 54, 15, 2688, 2010. 33. Plaza, I., Martín, L., Martin, S., Medrano, C., Mobile applications in an aging society: Status and trends. J. Syst. Softw., 84, 11, 1977, 2011. 34. Pang, Z., Chen, Q., Tian, J., Zheng, L., Dubrova, E., Ecosystem analysis in the design of open platform-based in-home healthcare terminals towards the internet-of-things. Proc. 2013, 15th Int. Conf. Adv. Commun. Technol. (ICACT), pp. 529–534, Pyeongchang, Korea, 2013. 35. Wei, Q., Zhu, S., Du, C., Study on key technologies of internet of things perceiving mine. Procedia Eng., 26, 2326, 2011. 36. Karakostas, B., A DNS architecture for the internet of things: A case study in transport logistics. Procedia Comput. Sci., 19, 594, 2013. 37. Plageras, A.P., Psannis, K.E., Stergiou, C., Wang., H., Gupta, B.B., Efficient IoT-based sensor BIG Data collection–processing and analysis in smart buildings. Future Gener. Comput. Syst., 82, 349, 2018. https://doi.org/10.1016/j.future.2017.09.082. 38. Qin, E., Long, Y., Zhang, C., Huang, L., Cloud computing and the internet of things: Technology innovation in automobile service, in: Lecture Notes in Computer Science book series (LNCS), Springer Berlin Heidelberg, vol. 8017, pp. 173–180, New York, USA, 2013. 39. Pang, Z., Chen, Q., Han, W., Zheng, L., Value-centric design of the internet-of-things solution for food supply chain: Value creation, sensor portfolio and information fusion. Inf. Syst. Front., 17, 2289, 2015. 40. Lee, C.K.M., Zhang, S.Z., Ng, K.K.H., Development of an industrial Internet of things suite for smart factory towards re-industrialization. Adv. Manuf., 5, 335, 2017. 41. Miorandi, D., Sicari, S., De Pellegrini, F., Chlamtac, I., Internet of things: Vision, applications and research challenges. Ad Hoc Netw., 10, 7, 1497, 2012. 42. Atzori, L., Iera, A., Morabito, G., The internet of things: A survey. Comput. Netw., 54, 15, 2787, 2010. 43. Wang, F., Ge, B., Zhang, L., Chen, Y., Xin, Y., Li, X., A system framework of security management in enterprise systems. Syst. Res. Behav. Sci., 30, 3, 287, 2013. 44. Xing, J.Y., Li, L., Bi, Z., Wilamowska-Korsak, M., Zhang, L., Operations research in service industries: A comprehensive review. Syst. Res. Behav. Sci., 30, 3, 300, 2013.

8 Sensor-Based Irrigation System: Introducing Technology in Agriculture Rohit Rastogi1*, Krishna Vir Singh1, Mihir Rai1, Kartik Sachdeva1, Tarun Yadav2 and Harshit Gupta1 Department of Computer Science and Engineering, ABES Engineering College, Ghaziabad, India 2 Department of Electronics and Communication Engineering, ABES Engineering College, Ghaziabad, India

1

Abstract

Agriculture sector has been the backbone of India. The contribution of the agriculture sector in the GDP has been much higher than the world average. Although, this has been falling gradually. This is primarily because of climate change, increasing water shortage and unawareness of farmers about optimal conditions and precision farming. IoT is advancing everywhere, leveraging the quality of products and increasing the production many folds. The agriculture sector has been lagging in being connected to IoT because of the current expensive solutions. In this paper, we introduce a concept for sector-based smart and precision farming most importantly cheap in cost using IoT which senses the weather and soil conditions of the farm-like soil moisture and temperature which is then used to irrigate the field accordingly. This is connected to the mobile app via the cloud service which can be used to view the data and control the actuators manually. Keywords:  IoT, agriculture, sensor & actuator, humidity, soil moisture, drip irrigation, RF module

8.1 Introduction Currently, most of the farmers flood the field with water. This causes a lot of wastage of water and sometimes excess, as irrigation is being followed by precipitation may damage the yield. The extra water that seeps down the ground also takes along a lot of essential nutrients with it. However, presently, farmers in India have great dependency and rely on non-technical and traditional irrigation methods in different times of the year for different period. Also, the time slag and demographical differences are also varying very fast in Indian continent, so accordingly, the moisture in soil and water requirement of crops also vary very much. It is not just the soil moisture but a lot of other factors that contribute in enhancing the yield, which are measured in this project [13]. *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (153–166) © 2021 Scrivener Publishing LLC

153

154  The Smart Cyber Ecosystem for Sustainable Development

8.1.1 Technology in Agriculture Real-time local weather monitoring prevents crop damage and is the first step in precision agriculture [1]. Appropriate irrigation is the most essential task in agriculture production to maximize yields, but there is growing evidence that optimized irrigation practices also lead to improved post-harvest product quality leading to reduced crop waste through the supply chain. Recent developments in soil sensors, wireless technologies, and application equipment provide a timely opportunity to develop a closed-loop system having potential of applying water variably across field crops on overhead irrigation systems. Along with this, new techniques that rapidly measure plant water status and improved soil moisture sensing technology will allow irrigation to be scheduled from plant responses rather than solely soil water availability.

8.1.2 Use and Need for Low-Cost Technology in Agriculture Deploying multiple sensor and actuator stations enable significant water conservation in open-field agriculture production, with a closely linked reduction in nutrient leaching and energy used for pumping. This project aims at making an analysis of the weather and soil conditions of an agriculture field and fulfilling the requirements of the same automatically for optimal crop yield using IoT minimizing human intervention.

8.2 Proposed System We propose a system as shown in Figure 8.1 to monitor the weather and soil conditions of the field and run actuators accordingly without human intervention. Multiple sensor and actuator stations [2] are deployed in the field. Each sensor station consists of an Arduino [3] microcontroller, moisture sensor, RF transmitter, relay, and a solenoid valve. Resistive Type Sensor is connected to it which measures the moisture content of the soil. The RF transmitter is used to send data to the central node. RF module has been preferred over the Wi-Fi module because it is not only cheap in price but also battery efficient [4]. The central node consists of a Raspberry Pi, Arduino, RF Transmitter, and a DHT Sensor [5]. From here, the data received from all the nodes is sent to the cloud. At the user end, a mobile app fetches the data in real-time from the cloud. This app is very useful to monitor if there is an undesired situation and to send necessary instructions to the central node to perform actions using actuators [6]. The farmer can monitor the farm conditions remotely anytime. The app has two modes: manual and automatic. In the manual mode, the app sends a notification alert whenever the soil moisture goes below the desired threshold. The farmer can manually turn on or off the irrigators of each sector independently directly from the mobile app. In the automatic mode, when the moisture comes below the set threshold, the controller irrigates the field with water by turning on the solenoid valve of that particular sector of the field until the desired moisture level is

Sensor-Based Irrigation System  155 Arduino

Arduino

Arduino

Wi-Fi

Internet

Cloud

Wi-Fi

Wi-Fi

Internet Raspberry Pi

Figure 8.1  Architecture of the system.

reached in that particular area [7]. Solenoid valves are laid in the field with each sensor station which supplies water to the field using a micro-irrigation system. These valves are directly connected to the central water pump (Figure 8.1). It is necessary to monitor the field humidity and temperature because in standard meteorological systems, information fetched from weather stations is entered into a calculation model. The data fed into it is received from a large number of stations, generally 1 system per 32-km resolution. Thus, this information is of a very large area and hence less precise. Placing more stations is difficult and costly [8]. Also, at many remote places, stations are not placed. Hence, we are providing a localized weather information of the farm to get precise and accurate readings which help in precision agriculture (Figure 8.2). This not only helps in optimal usage of water but also increases the yield of the crop by preventing the flowing of nutrients from the soil. The field humidity level and temperature is the most basic tool in doing precision farming, i.e., minimizing the use of resources. The data points produced can further be used for analytics which can contribute to research and increasing the crop yield. When the app is in automatic mode, it requires no human intervention and maintains the aforesaid conditions of the farm itself independently according to the parameters set.

156  The Smart Cyber Ecosystem for Sustainable Development

Raspberry Pi

Arduino

Arduino

Soil Moisture Sensor

Gas Sensor

DHT 11

Figure 8.2  Sequence tree for the whole system from center node to station nodes.

Figure 8.3  Connection diagram for each sensor station.

Arduino

Sensor-Based Irrigation System  157

8.3 Flow Chart Start

Read all sensor value

Temperature & Humidity sensor

Yes

Motor on

Moisture sensor

Sensor value > Threshold Value

Motor off

Raspberry pi

Cloud

Mobile app

Output

End

Process Chart 8.1  Flow chart for the system.

no

158  The Smart Cyber Ecosystem for Sustainable Development

8.4 Use Case Farmer Use Case Automatic

Irrigation Control

Statistics

Manual

Farmer

Sensing station

Add



Remove

Process Chart 8.2  Use case for the system.

8.5 System Modules 8.5.1 Raspberry Pi Raspberry Pi as shown in Figure 8.4 is a cheap credit-card sized central unit used in IoT projects. It is capable of doing everything which can be performed on a desktop computer. It operates on Raspbian OS installed on an SD Card. It has four USB ports and an HDMI port and multiple digital ports for connecting devices like controllers and sensors actuators [9].

8.5.2 Arduino Uno Arduino Uno as shown in Figure 8.5 is an open-source microcontroller board that uses the ATmega328P chip [3].

8.5.3 DHT 11 Humidity and Temperature Sensor DHT sensor as shown in Figure 8.6 is used to detect temperature and humidity for monitoring the weather conditions [5]. Temperature Range: 0°C–50°C Accuracy: ±2°C Humidity Range: 20% RH to 90% RH Humidity Accuracy: ±5% RH

Sensor-Based Irrigation System  159

Figure 8.4  Raspberry Pi Board, Halfacree Gareth, and Raspberry Pi 3 B + single-board computer (https:// upload.wikimedia.org/wikipedia/commons/b/b8/Raspberry_Pi_3_B%2B_%2839906370335%29.png).

Figure 8.5  Arduino Uno Development Board, Electronics SparkFun, and Arduino Uno Revision 3 (https:// hu.wikipedia.org/wiki/Arduino#/media/F%C3%A1jl:Arduino_Uno_-_R3.jpg).

Figure 8.6  DHT-11 Sensor, DHT-11 Sensor Pinout, features and datasheet, DHT11–temperature, and humidity sensor (https://components101.com/dht11-temperature-sensor).

160  The Smart Cyber Ecosystem for Sustainable Development

Warning Line

soil

Recommend Depth

Figure 8.7  Soil moisture sensor with the precautionary warning threshold for installation, Muzdrikah Fajar, calibration of capacitive soil moisture sensor (https://www.researchgate.net/figure/Capasitive-Soil-MoistureSensor-SKUSEN0193_fig1_329492087).

8.5.4 Soil Moisture Sensor Soil moisture sensor as shown in Figure 8.7 is used to capture the moisture content of the soil by measuring the voltage in soil using soil moisture as the dielectric. It then can convert the voltage readings to percentage.

8.5.5 Solenoid Valve It is an electric valve working on DC. It opens when the current is supplied to it and gets closed when no current is present. It controls the flow of water. It is generally used with microcontroller boards with relays attached to them.

8.5.6 Drip Irrigation Kit Drip irrigation kit comes with all the parts to assemble the drip irrigation system at any plant unit. Comprises flexible pipes with drip controller heads and all the required assembly units too that helps in structuring the unit from plants to water supply.

8.5.7 433 MHz RF Module This is a transmitter and receiver module used to wirelessly connect the central and the Arduino stations using radio frequency waves working at 433 Mhz. This works for large distances with minimal noise in the channel [4].

8.5.8 Mobile Application This application developed in Flutter [10]. It can be used to monitor the real time and historical data of all the sensors present in the farm [6]. It directly fetches the data from the

Sensor-Based Irrigation System  161 cloud. It can be operated in two modes: automatic mode—in this mode, the actuators run automatically when the desired conditions are dissatisfied without any human interventions; manual mode—in this mode, the separate actuators can be controlled manually [11]. The demonstration snapshot 1 shows that how the UI of the mobile app looks like, from this page a user can control the actuators (both automatically and manually) can view data of the sensing stations as shown in demonstration snapshot 2 and can add the sensing station.

8.5.9 Testing Phase The following images are from the testing phase of the P.o.C. Some of the on-ground demonstration and calibration was done in the phase, and according to the result, improvements were discussed for the next iteration of the prototype. Testing Figures 8.1 (a), 8.1 (b) and 8.2 (a) and 8.2 (b).

Testing Figure 8.1  (a) Installation of the device in the field for real-time data values.

Testing Figure 8.1  (b) Device installed and delivering the data to the center node in the lab.

162  The Smart Cyber Ecosystem for Sustainable Development

Testing Figure 8.2  (a) Test application interaction UI.

Testing Figure 8.2  (b) Fetched testing data screen.

8.6 Limitations The system works well when the surface is even and moisture spread on it evenly. For this, the irrigation system also needs to ensure uniform distribution. It will not be successful in hilly areas. The sensor needs to be calibrated for different soil types as soil conditions vary from one area to another. The sensing stations need a continuous power source which brings the need to lay wires in the field. The sensors used are point sensors and hence need to be densely deployed.

8.7 Suggestions The sensing stations can be connected with rechargeable batteries so that there is no need to lay wires. The sensors can be made self-calibrating so that there is no need to manually calibrate them at the time of every installation.

8.8 Future Scope This application can be advanced further when deployed in greenhouses where we can sense the amount of light, humidity and wind speed. Then, we can use fans to give optimal air flow, blow off the toxic gases or the excess moisture in the greenhouse [6]. The data acquired from the sensors can be used for a lot of research purposes and advancements in the agriculture sector.

Sensor-Based Irrigation System  163

8.9 Conclusion Irrigation is a necessary and inevitable task in farming. On the other hand, water resources are getting depleted day by day. This paper proposed a successful system of sector-based smart irrigation systems, which helps in improving the yields by automatically maintaining the soil moisture and also collecting the soil and weather data. This will not only automate the cumbersome task of watering the fields, thus improving the lifestyles of farmers but will also contribute to water conservation. This project can be improved further by giving more precise and accurate predictions. Increasing the frequency of data capturing can help in making better analysis and improving productivity but at the same time will deplete the battery life of the stations deployed in the field and will also need more storage.

Acknowledgement First and foremost, our team would like to thank the almighty for everything that we accomplished till now. With this, we would like to extend our gratitude toward Prof. Shailesh Tiwari, Director, ABES-Engineering College, Ghaziabad, U.P., India. He has been supportive since the first day we joined the college, and with his support and opportunity, our team is able to work any time in the lab and get this project from scratch to finish. We would like to offer our gratitude toward Mr. Amit Goyal, Chief Executive Officer, DataRitz Technology, and Mr. Gaurav Kansal, Chief Operating Officer, DataRitz Technology, HoD, CBSE, ABES-EC. With their constant support and guidance, this project made tremendous improvement from the day it was just an idea. Both of them were there at the start when it came to helping in any sort. We would like to thank Dr. Pradeep Kr. Singh, HoD, CSE Dept., who always believed in our team and helped us in the project in every way possible. He encouraged us to pursue project work even during examinations and helped us manage time with academics. Last but not least, we would like to pay our gratitudes toward the team of management, our colleagues, and the whole staff of the ABES-EC for helping our team during different times. It is with great pleasure we offer gratitude to all and anyone not mentioned in this acknowledgment, gratitude is due.

References 1. Nayyar, V.P.A., Smart farming: IoT based smart sensors agriculture stick for live temperature and moisture monitoring using Arduino, cloud computing & solar technology. The International Conference on Communication and Computing Systems (ICCCS-2016), 2016. 2. Navulur, A. S. C. S. S. M. N. G. P. Sr., Agricultural Management through Wireless Sensors and Internet of Things. Int. J. Electr. Comput. Eng. (IJECE), 7, 2017. 3. Badamasi, Y.A., The working principle of an Arduino. 11th International Conference on Electronics, Computer and Computation (ICECCO), Abuja, 2014. 4. Ahmed, F. e. a., 433 MHz (Wireless RF) communication between two Arduino UNO. Am. J. Eng. Res. (AJER), 5.10, 10, 358–362, 2016. 5. Tianlong, N., Application of Single Bus Sensor DHT11 in Temperature Humidity Measure and Control System [J]. Microcontrollers Embedded Syst., 026, 2010.

164  The Smart Cyber Ecosystem for Sustainable Development 6. Siwakorn Jindarat, P.W., Smart Farm Monitoring Using Raspberry Pi and Arduino. IEEE 2015 International Conference on Computer, Communication, and Control Technology, 2015. 7. J. S. J. A. M. V. M. A. F. S. Zamora-Izquierdo, M.A., Smart farming IoT platform based on edge and cloud computing, Biosystems Engineering, M.A. Zamora-Izquierdo, et al., (Eds.) Special Issue of Bio Systems Engineering, Science Direct, 177, 4–17, 2019. 8. R. K. A. T. J. S. Sabharwal, N., A Low cost Zigbee based automatic wireless weather station with GUI and web hosting facility. ICRTEDC, 2014. 9. Vujović, V. a. M. M., Raspberry Pi as a Wireless Sensor node: Performances and constraints. 37th International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO), IEEE, Opatija, 2014. 10. Wu, W., Hello Flutter, in: Beginning App Development with Flutter, pp. 3–8, Academic Press, Berkeley, CA, 2018. 11. M. B. A. N. S. D. Patil, A., Smart Farming using Arduino and Data Mining. 2016 International Conference on Computing for Sustainable Global Development (INDIACom), 2016. 12. Gärdenäs, A. I. e. a., Two-dimensional modeling of nitrate leaching for various fertigation scenarios under micro-irrigation. Agric. Water Manage., 74, 3, 219–242, 2005. 13. National Portal of India, in: Topic-Agriculture, 2020, https://www.india.gov.in/topics/agriculture.

Suggested Additional Readings • Sensor-based irrigation management of soilless basil using a new smart irrigation system: Effects of set-point on plant physiological responses and crop performance https://www.sciencedirect.com/science/article/abs/pii/S0378377418301057 • Sensor based Automated Irrigation System with IoT: A Technical Review http://ijcsit.com/docs/Volume%206/vol6issue06/ijcsit20150606104.pdf • Sensor Based Automated Irrigation System https://www.ijert.org/research/sensor-based-automated-irrigation-systemIJERTV4IS050076.pdf • Sensor Based Automatic Irrigation Management System https://www.ijcit.com/archives/volume4/issue3/Paper040304.pdf • 3 Ways to Automatic Plan Irrigation System using Microcontroller https://www.elprocus.com/microcontroller-based-automatic-irrigation-system/

Key Terms and Definitions IoT: The Internet of Things (IoT) is a system of interrelated computing devices, mechanical and digital machines provided with unique identifiers (UIDs) and the ability to transfer data over a network without requiring human-tohuman or human-to-computer interaction. Agriculture: Agriculture is the science and art of cultivating plants and livestock. Agriculture was the key development in the rise of sedentary human civilization, whereby farming of domesticated species created food surpluses that enabled people to live in cities. Artificial Intelligence: In computer science, artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines,

Sensor-Based Irrigation System  165 in contrast to the natural intelligence displayed by humans and animals. Colloquially, the term “artificial intelligence” is often used to describe machines (or computers) that mimic “cognitive” functions that humans associate with the human mind, such as “learning” and “problem solving”. Machine Learning: Machine learning (ML) is the study of computer algorithms that improve automatically through experience. It is seen as a subset of artificial intelligence. ML algorithms build a mathematical model based on sample data, known as “training data”, in order to make predictions or decisions without being explicitly programmed to do so. Arduino: Arduino is an open-source hardware and software company, project, and user community that designs and manufactures single-board microcontrollers and microcontroller kits for building digital devices. Its products are licensed under the GNU Lesser General Public License (LGPL) or the GNU General Public License (GPL), permitting the manufacture of Arduino boards and software distribution by anyone. Arduino boards are available commercially in preassembled form or as do-it-yourself (DIY) kits.

Appendix Sr. no.

Humidity

Moisture

Temperature

Gas

Time

1

59

31

24

457

Tue Aug 13 11:52:23 2019

2

60

32

25

454

Tue Aug 13 11:52:34 2019

3

60

32

24

452

Tue Aug 13 11:52:44 2019

4

60

33

24

451

Tue Aug 13 11:53:10 2019

5

60

33

24

450

Tue Aug 13 11:53:52 2019

6

65

39

27

354

Wed Aug 14 14:18:20 2019

7

64

38

27

319

Wed Aug 14 14:20:08 2019

8

64

38

37

318

Wed Aug 14 14:20:26 2019

9

64

43

27

357

Wed Aug 14 14:20:41 2019

10

63

44

27

362

Wed Aug 14 14:20:54 2019

11

62

38

27

307

Wed Aug 14 14:22:46 2019

12

59

43

27

319

Wed Aug 14 14:25:14 2019

13

60

43

27

323

Wed Aug 14 14:25:28 2019

14

60

43

27

321

Wed Aug 14 14:25:33 2019

15

56

39

26

243

Wed Aug 14 14:33:27 2019

16

56

39

26

242

Wed Aug 14 14:34:04 2019

17

56

39

26

241

Wed Aug 14 14:34:35 2019

18

56

39

26

240

Wed Aug 14 14:35:00 2019

19

55

39

26

239

Wed Aug 14 14:35:11 2019

20

55

39

26

238

Wed Aug 14 14:35:22 2019

166  The Smart Cyber Ecosystem for Sustainable Development

Example Code

A1  Serial input from the input device (Here Arduino) through USB.

A2  Organizing the data in returnable functions to maintain modularity.

9 Artificial Intelligence: An Imaginary World of Machine Bharat C. Patel1*, Manish M. Kaysth2 and Tejaskumar R. Ghadiyali2 Smt. Tanuben and Dr. Manubhai Trivedi College of Information Science, Wadia Women’s College Campus, Athwaline, Surat, Gujarat, India 2 Udhna Citizen Commerce College & SPB College of Business Administration & SDHG College of BCA & IT, Surat, Gujarat, India 1

Abstract

Artificial Intelligence (AI) is the process of simulating the thoughts of human being into a machine so that the machine can be capable to demonstrate traits associated with a human mind for example learning and problem-solving. A machine is also capable to perform various tasks which are similar to the task being carried out by human being such as a machine can learn from experience; it alters new inputs by itself and executes it. Therefore, sometimes, AI is also known as machine intelligence. AI is a technology that has been moved from science fiction to business fact. During the journey of this chapter, we will discuss the history of AI, the fundamental concept and types of AI, the future trends in AI, challenges in AI, and various application areas within the different sectors. Finally, we will discuss the practical demonstration of an application that translate textual image into editable digital form using neural network. Keywords:  Artificial Intelligence, components of AI, challenges and trends in AI, application area of AI

9.1 The Dawn of Artificial Intelligence In computer science, Artificial Intelligence (AI) became a very exciting and attractive field for the researchers. At present, we are moving on extremely interesting era of the AI. Over the past decade, AI has made a remarkable progress in the field of computer science. The awareness of non-living objects were gifted with intelligence has been trying around since earliest era. The inspiration of AI was established very long time ago. In the olden Egypt, the statues of gods animated are prepared by engineers with the help of priests. Hundred years ago, the philosophers have illustrated the process of human thinking in the form of symbols using the logic and tools; and established the foundation for the concept of AI such as general knowledge intelligence. *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (167–184) © 2021 Scrivener Publishing LLC

167

168  The Smart Cyber Ecosystem for Sustainable Development The foundation work of AI would deliver acceleration to the modern computers within the interval of very last 19th and initial half of the 20th century. The primary design used for a programmable machine would be invented by two mathematicians Charles Babbage and Augusta Ada Byron at Cambridge University in 1836. The architecture designed for the stored-program would enable program and data to be kept in the computer’s memory that was first imagined by Princeton mathematician John Von Neumann in the 1940s. In 1943, the first initiative was attempted by Warren McCulloch and Walter Pitts to design a model, known as neural network (NN), to know the operation of neurons [1].

AI is initiated within the year 1950 with the vision to resolve complex problems of mathematics and make a “thinking machines” or “intelligent machine”. The British mathe-

matician Alan Turing, a code-breaker of Second World War, was the pioneer of AI. In 1950, he predicted that by the end of century the machine would be ready to think like human can do. Alan had developed a system, named “Turing Test”, in order to determine the intelligence of a computer [2]. John McCarthy proposed a term AI in 1956 at the Dartmouth conference. He was the person who invented a programming language for AI referred to as LISP, stands for LISt Processing, within the year 1958 and demonstration of the first running AI program at Carnegie Mellon University. John McCarthy has defined the meaning of AI as it is a combination of science and engineering to develop intelligent programs in order to make intelligent machines. In 1965, Joseph Weizenbaum developed a first realistic chatbot called ELIZA. It is a natural language program that handles dialogue on any topic, similar in concept to today’s chatbots. In the initial 20 years, AI was the foremost successful approach that was resulting in immense excitement and significant government funding. But in real-world settings AI did not achieve its outcomes within the 1970s. Artificial NNs were also struggling, funding dried up, research slowed and also the AI community gets smaller. In the 1980s, Edward Feigenbaum created an expert system which emulates decisions of human experts. In 1997, IBM has developed a computer chess program named “Deep Blue” which beats Garry Kasparov, a Russian chess Grandmaster, and get the honor of first computer program to defeat a world champion. After 14 years (in 2011), IBM launched question-answering computer system, known as Watson, which is able to give the answer of the questions ask in human natural language. It overthrows in the contest with two Jeopardy greatest champions of US game show. In 2002, iRobot launched a vacuum cleaner, known as Roomba, which autonomously navigate and remove obstacles appear on its way. In 2009, Google launched the foremost self-driving car using the concept of AI to handle urban situation [3]. Personal assistants like Siri (2010), Google Now (2012), and cortana (2014) were developed by Apple, Google and Microsoft, respectively. They employ speech recognition techniques to answer the questions and perform simple task. In 2014, Ian Goodfellow comes up with Generative Adversarial Network (GAN) deep learning frameworks within which two models, a generative model and a discriminative model, which are trained simultaneously and generates output [4]. The Deep Mind Technology has developed AlphaGo which was later acquired by Google. In 2016, the AlphaGo beats the professional board game Go player Lee Sedol and manifest a most important objective in the growth of intelligent machines. Figure 9.1 shows brief sketch of history of AI.

AI: An Imaginary World of Machine  169

Edward Feigenbaum creates expert system

AlphaGo beats the board game Go player Lee Sedol

Watson – defeated two of Jeopardy’s champions of US game show

2016 2014

2011

2009

Roomba – an autonomous vacuum cleaner launches by iRobot

Personal Assistant like Siri, Google Now, Cortana created 2011–2014

Google developed self-driving car

2002

1980

1955 Al is coined during a conference devoted to the topic

Deep Blue – program developed by IBM

1997

ELIZA – first realistic chatbox any topic

1965

1950

Turing Test-launch by Alan Turingtesting intelligence of machine

GAN – deep learning framework by Ian Goodfellow

Figure 9.1  Timeline evolution of Artificial Intelligence.

In the last few years, the most recent inclination of researchers are focusing on the incredible subfield of Machine Learning (ML) known as Deep Learning (DL); it is one kind of biologically inspired NN, that has the capability to manage large amount of data, enormous computational power, and faster execution speed of computers.

9.2 Introduction In the current era, AI is rapidly advancing technology and plays vital role within the field of computer science. Traditionally, AI is an artificial creation of intelligence technique, similar to human intelligence, which is able to study the reason, gather information, communicate, recognize, and manipulate the expected language [5]. AI is the process of simulates thoughts of human beings which applies to any machine that demonstrate traits related to somebody’s mind like learning and problem-solving. A machine is capable to perform a task similar to human beings such as learn from the experience, adjust to new inputs and execute, therefore sometimes, it is also referred to as machine intelligence. AI is a technology that has been gradually transferred from science fiction to business fact. Dictionaries define intelligence as the capability to obtain, understand, and apply the understanding, or the ability to implement idea and reason. Obviously, intelligence is too much than this [6]. Roughly say the meaning of AI means non-natural intelligence. We will make a machine capable to incorporate this non-natural intelligence in order to exhibit like intelligence of a human being. In other words, AI may be a technique that allows machines to gain intelligence similar to human beings and work efficiently using algorithm provided to the system. In the current scenario, two more terminologies like ML and DL play a vital role within the mind of the researchers of computer science. We can visualize the relationship among all three technologies are depicted as shown in Figure 9.2. As shown in Figure 9.2, AI may be a superset of ML and DL but one of the groups of researchers’ believed that this relationship is not accepted universally by all. There is

170  The Smart Cyber Ecosystem for Sustainable Development Artificial Intelligence Machine Learning Deep Learning

Figure 9.2  Relationship between AI, ML, and DP.

another group that believed that ML is not a subset of AI, but both AI and ML with some overlap between them. The main thing about these technologies is that they are algorithms and used for solving social and business problems. AI could be a very huge area of computer science that covers technologies like ML and DL. It could be a technology in which a machine, especially a computer system, has been trained in order to acquire the intelligence of the human process. ML consists of collection of algorithm that uses the data to carry out work like prediction and classification competently. The machine acquires learning ability by means of supplementary models and data provided to it. The algorithm is called learning because with the help of training and past experience, it improves the correctness of classification task. The concept of DL came into existence first time in 2006 as a new field of research in ML. Initially, it is absolutely called hierarchical learning at the [7, 8] and typically it contains various research fields related to pattern recognition. DL emulates the functioning of the human brain to unravel problems. DL is considered as a way to increase the outcome and effective processing time. Nowadays, there are many applications available that includes AI, ML, and DL. One of the applications is Apple’s SIRI virtual assistant, which is able to process the human speech and language, enter it into a Google’s search engine, and return recognized output to us.

9.3 Components of AI AI can be considered as the big field of describing systems having an ability to think. As illustrated in Figure 9.3, there are four main subsets of AI such as Machine Reasoning, Natural Language Processing (NLP), Automated Planning, and ML [9, 10].

9.3.1 Machine Reasoning Machine reasoning consent to a system to prepare a hypothesis based on data. It helps machine to fill in the gaps when there is incomplete data, i.e., it helps to create a sense of related data. For example, a system has enough data and you asked questions, the system would be capable to telling you the answer of that question.

AI: An Imaginary World of Machine  171 Artificial Intelligence Reasoning

Natural Language Processing

Planning

Machine Learning Supervised Learning

Reinforcement Learning

Unsupervised Learning

Deep Learning -Neural Network

Figure 9.3  Components of Artificial Intelligence.

9.3.2 Natural Language Processing This subfield of AI can described the relation between computers and human languages. NLP has a potential to guide computers in order to recognize both written text and human speech. In a very layman term, NLP is nothing but the processing of human language using computer program with the intention to understand written text as well as human languages. It picks up the meaning of unstructured data from papers or user’s communication facts. Therefore, NLP is the fundamental technique that can understand text and language spoken by human beings. Additionally, it also allows permission to non-technical persons to communicate with modern technologies. As an example, instead of writing tedious task of program code, NLP can assist users in such a way that the user has to just ask questions about complex data sets through the system. Categorization, catalogs, dictionaries, ontologies, tapping, and language models are the necessary tools used for NLP.

9.3.3 Automated Planning It is the capability for an intelligent system to carry out task separately and flexibly to create a series of events to accomplish a final objective. Typically, in the preprogrammed decisionmaking procedure, the order of process execution goes from processes P1 to P2 to P3 and at last come to ultimate result. Obviously, the automated planning process is hard to implement and the system require making a decision in order to handle the condition encountered around the challenge at hand.

9.3.4 Machine Learning The better correctness of predictive model is achieved by use of this technique. According to the business problems, like type and volume, it has four different learning algorithms such as (i) Supervised learning, (ii) Unsupervised learning, (iii) Reinforcement learning, and (iv) NNs and DL.

172  The Smart Cyber Ecosystem for Sustainable Development

9.4 Types of Artificial Intelligence AI is the science of being compensated machines to copy the activities of humans. Moreover, as we discussed in the previous section that ML is a subfield of AI which enables machine to make judgment through the data given to it. In addition, DL is a component of ML that solve complicated problems with the concept of NNs. So to submit up, all three technologies such as AI, ML, and DL are interconnected fields. Both ML and DL are facilitating to resolve data-driven problem by providing a collection of algorithm and NN to AI and it is not restricted to only ML and DL. There are a variety of fields like computer vision, NLP, robotics, object detection, expert system, etc., are included in AI. AI can be structured along three evolutionary things: (i) Artificial Narrow Intelligence (ANI), (ii) Artificial General Intelligence (AGI), and (iii) Artificial Super Intelligence (ASI) [11]. In the following sections, we will discuss the concepts of all three categories with their application in the different fields.

9.4.1 Artificial Narrow Intelligence A process of replicate intelligence and behavior of human being using the machine which is confined to a limited range is called ANI. It is also known as weak AI and a task is performed by it within a predefined and predetermined range. The AI is applying to the machine so that it works efficiently and fulfills the task in specific domain. The machine is able to play chess with human, able to forecast weather report and many more things are examples of ANI. ANI can accomplish specific task fabulously, with the help of advanced algorithms, DL, and various techniques depending on the usage. Every type of machine intelligence designed for specific task is considered as narrow AI. Some of the virtual assistants like Siri by Apple, Cortana by Microsoft, Alexa by Amazon, Google translate, and other NLP tools are satisfied the criteria of narrow AI. The term narrow indicates that the system is designed to perform specific task very well but they cannot have human-like intelligence such as self-awareness, consciousness, and genuine intelligence to match human intelligence. For example, as soon as we communicate with SIRI, it is ready to answer the questions that are implemented in the system design to do so. The apple’s virtual assistance, SIRI, is able to identify the human speech entered into a search engine and come back with efficient outcome. It is unconscious machine and not able to answer the question outside the domain. Similarly, a machine can beat the world chess champion, but that is the only thing it can do. Humans have characteristics like to analysis the environment around them, responsive in nature and give reply as sensitive-driven on the basis of circumstances, whereas AI machine present around us has no variability or flexibility as human beings have. Some of the applications listed below are based on weak AI or ANI: • Face verification at Apple iPhone: In 2017, Apple launched new Face ID technology on the iPhone X. It has a capability to form 3D map (of more than 30,000 invisible dots) of the face using the graphics technique. It uses A11 Bionic chip to process data and recognize the changes in your look with the

AI: An Imaginary World of Machine  173

•

•

•

•

help of ML algorithm. The recognition algorithm is so intelligence that even glasses on face, wear a hat, grow a beard, and wild makeup will not make fool to Face ID [12, 13]. Autopilot features at Tesla: Tesla Autopilot driving technology makes huge evolution on autonomous driving technology. It provides driver an automatic access to the information that is used to control its actions. It handles functionalities like navigate and transport using AI together with accident prevention technology called Advanced Driver Assistance Systems (ADAS). It activates emergency steering and braking system of the autopilot technology. Tesla Model S and Model X have features like AutoSteer, AutoLane, Automatic Emergency Steering, Side Collision Warning, and AutoPark that would reduce most tedious and dangerous aspects of road travel [14]. The social humanoid, Sophia: David Hanson developed Sophia, the social humanoid, at Hanson Robotics in 2015. Some activities that robots with the computer science perform speech recognition, learning, problem finding, and planning using the concept of AI [15]. Finding the optimal path through Google Maps: The Google Maps team launched an excellent feature called Street View in 2007. Using AI and other tools, Google offers functionalities like street View, real-time traffic conditions, street maps, route planning for traveling by foot, car, bicycle and air, satellite imagery, public transportation, and aerial photography [16]. Virtual assistance Alexa: An AI-based virtual assistance, Alexa, is launched by Amazon in 2014. It has a limited predefined range of functions. It has no genuine intelligence, self-awareness, and narrowly defined specific problems [17].

9.4.2 Artificial General Intelligence As we discussed in ANI that it is designed to do any specific task, whereas AGI is designed to learn to anything. It is some level far better than ANI and therefore; AGI is also known as strong AI. Furthermore, the machine has imaginary intelligence and has a capability to understand or gain knowledge of any intellectual task as smart as a human being can. The machine has strong processing units that can perform high level contest but in some circumstances it cannot think and reasoning similar to human being. AI is also used in the Hollywood movies, like “Her” or “sci-fi”, in which the actors interact with conscious and sentient machines. At the present time, the advancement in computer peripheral manufacturing industries, computer machines process data much quicker than human being can do. Although human beings have the characteristics like ability to imagine conceptually, manipulate, and come out with opinion and recollections to make knowledgeable decision. This type of intelligence makes human more superior to machine, besides replicating this intelligence in a machine is very difficult. AGI is likely to be capable to find out reason, determine problems and solve them, take a decision in ambiguity, plan, discover, and put together past information in decision-making. To get exact human-like intelligence, the machines have to become ready to equip with knowledgeable awareness.

174  The Smart Cyber Ecosystem for Sustainable Development Let us visualize some of the examples that can handle multiple tasks depending upon the problem in hand wherein a little bit use of AGI is used. ✓✓ A Philips screwdriver can look at the problem and transform into appropriate right tool, like a saw or a paint brush or a tape measure, to solve the problem. ✓✓ Similarly, let us consider software application that can be adjust as per the environment in which you are working on; for example, PowerPoint changes into Photoshop, then Excel, WhatsApp, and SAP to perform different type of tasks. ✓✓ A future robot could be transform into another form as per the problem face by itself.

9.4.3 Artificial Super Intelligence ASI is a term touching on the time when the potential of computers will surpass humans. That means, the machine has its own intelligence and thinking power go beyond the human intelligence. ASI will beat human intelligence in all phases like from imagination, general knowledge, and analytical process. The computers would be able to show intelligence which is not seen up till now. Many researchers and philosophers will be worried about such kind of AI and its vulnerability. ASI is presently seen as theoretical condition since detecting movies and science fiction tasks where a machine takes over the world. However, many researchers are forecasting that ASI takes over the world by the year 2040. Though, we are many years away from ASI but some researchers predicting that to move from AGI to ASI will be a little one. Nobody can actually aware about when the first conscious or responsive computer life is started. However, like narrow AI progressively more elegant and skillful, we will believe that future of AI is regulating by both men and machine. We can visualize some of the tiny application of ASI as follows: • Spotting an eight-planet solar system which is 2,500 light year away. • Composing sonnet and poem. • In financial sector, JPMorgan Chase’s contract intelligence (COiN) platform utilize AI, ML, and image recognition software to analyze legal documents and extract important data points and clauses in the matter of seconds. At this moment, manually reviewing 12,000 agreements takes over 36,000 hours but AI would be able to do that in the matter of seconds. • In the healthcare, IBM is one of the pioneers that have developed AI software especially for medicine. More than 230 healthcare organizations world-wide use IBM Watson technology. In 2016, IBM Watson AI technology is able to cross-reference 20 million on collegy records and correctly diagnosis rare leukemia condenses in the patients. • Google’s AI eye doctor can examine retina scans and identify a condition known as diabetic retinopathy which causes blindness. • Facebook utilizes the concepts of ML and DL for face verification to detect facial attributes and label your associates.

AI: An Imaginary World of Machine  175 • Twitter makes use of AI to discover negative text and panic speech inside the tweets. It utilizes the features of ML, DL, and NLP to filter fansive context. The software industries find out and exposed prohibited more than 3 lakhs terrorist-linked accounts, out of that 95% accounts would be accessed by the computer with the help of AI. • Google Predictive Search (GPS) is the most famous AI application. When you start writing a search term, Google offers you recommended terms to choose from that means an AI in action. Google collect information like your location, age, and other personal details with the help of predictive search. • Self-driving car: Autonomous driving car is manufactured by Tesla utilize the AI methods like DL, image detection and computer vision. It automatically detects objects or any obstacles on its way and run without the help of human being. It must also learn turn about how AI is implemented in Tesla selfdriving car and auto pilot features. ASI incorporates a capability in specific areas where computer intelligence is better than the human’s one whereas ASI will have the aptitude to be better not just in an exceedingly few areas but better in most of the areas. The computers are becoming more intelligent by learning and that is why they do not learn from humans. They are beginning to get the knowledge by themselves, attempting to alternative approach and assessing them. Since AI is jumping so fast that in a few years, it takes every aspect of our life. However, some of the questions arise in our mind that, Is it possible that in the future AI might confine our lives? This will be unfortunately take place within the next 20 years, even though a few scientists and futurists think on this transition to ASI will happen in the fourth decade of this century.

9.5 Application Area of AI AI has geared up its approach into a large range of domains. Some of application areas that is surrounding along the AI concept are as follows. ✓✓ In the healthcare sector, the system extracts details information about patient from the available data sources to prepare assumptions. ML helps to healthcare application for quicker and superior diagnoses than human being. In the healthcare, IBM is one of the pioneers that have developed AI software especially for medicine is the Watson. It helps patients as well as healthcare customers in various activities like to find out medical details, appointments to be scheduled, supplementary administrative processes with the help of virtual health assistant and chatbots facilities included in the Watson [18]. ✓✓ In the business, ML algorithms are incorporated into Customer Relationship Management (CRM) platforms which help customer to provide services like seek detailed information using chatbots [19]. ✓✓ In education, AI can automate grading, giving educators extra time. It can evaluate students and adjust their needs, helping them in work on their own pace. AI tutors continuously observe the activity of students and check whether they are on the right way or not.

176  The Smart Cyber Ecosystem for Sustainable Development ✓✓ In the manufacturing industries, a daily routine task perform by worker is substitute by robots at one time programming. For example, the industrial cobots (Smaller) are the multitasking robots that work together with humans and carry out accountability for further task in storehouse and other workspaces. ✓✓ In transport sector, the fundamental role of AI is to operating autonomous vehicles. Artificial NN is useful to perform different activities in transportation like to control traffic, forecast flight waits and put together marine transport secure and well-organized [20]. ✓✓ Different organizations employ ML in the application known as Security Information and Event Management (SIEM) to discover abnormality and recognize mistrustful events that specify threats [21].

9.6 Challenges in Artificial Intelligence AI is a debatable subject, where some would say that it is excellent in businesses; whereas for someone, it is a technology that causes hazard to survival of human kind. In many of the sector, AI has altered our wealth either directly or indirectly. In future, there may be chances that the AI would have taken over several most important routine tasks. It may be carry away human being to lose their jobs by machines. Therefore, it is necessary to make attentions on some of the sectors where AI faces challenges given below: ✓✓ Decision-Making: AI agent takes critical decisions of any task in the AI ranging from self-driving cars to managing insurance payouts. But sometime, a judgment taken by the intelligence algorithm is not often confidential by the things like corporate. ML makes this phenomenon too difficult because the underline programming judgment of the system is not clear to the programmer as well as sometime difficult to know. This has inferences for the formation of ML systems, but more significant matter is that its secure operation and creditability. It is required to know that why a self-driving car not only to confirm the technical mechanism to choose a particular actions, but it also resolves problem in the case of an accident [22]. ✓✓ Social and Economic Impact: It is examined with the purpose that AI technologies will take revolution in finance by boosting in the productivity. Moreover, the machines have capability to carry out new tasks like highly developed robots, automatic cars, or skillful virtual assistants would give support to the people in the daily routine tasks. As per consumer’s viewpoint, automation implies better efficiency and cheaper products. Obviously, AI will also produce new vacancy or increase demand in particular sectors but automation may impact the separation of labor on a universal scale. In the last few decades, we have observed that many small financial sectors in production and services would be transferred from developed to emerging economies due to reasonably minor worker or material prices [23]. ✓✓ High Expectation: The major problem would be the people have extremely high hope with AI. Generally, people do not have thorough understanding of

AI: An Imaginary World of Machine  177 working of AI and their expectations are extremely high even though some of the operations are not possible. Humans have an expectation to predict more from something that is going to be developed and the generated result is also outstanding. However, compare to any other technology, AI has some drawbacks related with it. Business professionals of different industries are further confident and expected that future decades would be rule by AI, but there are several barriers that create major hurdles for AI [24]. ✓✓ Software Malfunction: It is obvious that neither technology nor person is an ideal. There is a break down in software or hardware; it is not easy to make anyone responsible for incorrect or flawed. Conversely, responsibilities performed by humans can be easily outlined. However, with machines and integral algorithms informed, it is not easy to blame someone or to find the cause of either software or hardware crash. ✓✓ Investment: Another great confront of AI is that not every industry owners or director are ready to expense on it. The configuration and implement of AI requires very expensive resources; as a result, neither every business owner nor organization can spend money on it or can take a risk in their business.

9.7 Future Trends in Artificial Intelligence The development of ML and AI related technologies will have a long drive in coming years. Well-known companies like IBM and Apple are giving more authority to research and development of AI. Therefore, they will diminish the gap between AI as well as consumers. Furthermore, some of the AI trends in next coming future will be described in this section. We have enlisted following trends of AI to understand the significance of latest technologies: ✓✓ Artificial Intelligence and Internet of Things (IoT): IoT is an enormous idea that consists of collection of software, sensors, and other technologies that are interrelated with each other to form a network. It also contains huge amount of data storage and data processing capabilities by means of the Internet. Last 15–20 years back, no one can have thought about virtual video chat with their friends as well as family members reside in other county. Nowadays, it is a universal thing which is possible because of technology getting cheaper, and the devices materialized with new improve and better potentials. IoT makes possible that everybody can carry out different transactions like transfer emails or money and booking ticket or cab using one tap on the mobile [25]. A well furnish IoT system utilize AI and ML that helps to manage huge volume of unstructured data. The digital things like sensors and chips generate a huge opportunity in tapping important data, running analytics and we can make up to date and better decisions with those. In the current era, we are living in the world where the control in our Smartphone is much more than the control in our homes which is possible by AI [26]. Few samples of available IoT services with the working of AI at the back of them are Voice assistants, Robots, Smart Devices and Industrial IoT.

178  The Smart Cyber Ecosystem for Sustainable Development ✓✓ Artificial Intelligence and Cloud Computing: Google, Amazon, Microsoft and IBM are the different cloud providers would put together of AI potential in cloud computing. AI and Cloud Computing are collectively called Intelligent Cloud Computing. Both cloud computing and AI has massive advantages. Cloud makes possible a digitally connected world and decreases the workload of controlling heavy hardware. AI makes possible to simplify the enormous facilities of cloud computing and furnish incredible control to the businesses. They offer ML platform of cloud and cloud services of AI like powerful text analysis and speech recognition, intelligent language, and knowledge and many more [27, 28]. The purpose of cloud computing is to make appropriate utilization of computer peripherals, i.e., hardware and software, that perform services on internet. Cloud computing give remote services with a user’s data, software and computation [29]. ✓✓ Facial Recognition (FR): This technology is observed as the outlook of AI because of its enormous reputation and will give guarantees to huge development in 2020 and in future. FR is an AI-based application that facilitates to recognize a person with their facial geometry or pattern. The year 2020 would be viewed as a development of FR tools with more dependability and better accurateness. Let us consider two latest examples like DeepFace program of Facebook and FR in the IPhoneX are based on the concept of FR technology. Recently, we are going in the direction of making everything personalized, whether advertising or shopping and so on. Therefore, FR technology would be proposed immense support in biometric identification. This technology has also made impact in the healthcare sector also for consequent clinical examinations and medical analytical methods. Openwater is an imaging technology which is able to reading images from human brains; it is to be expected have a bright future [30, 31]. ✓✓ Automated Machine Learning (AutoML): AutoML is a fast growing research field within the computer science that has the potential to help non-experts who want to learn ML themselves. The purpose of automated techniques is to check the status of the person in specific domain of ML within the small amount of instant. But, the foremost drawback of AutoML at this moment is not able to work powerfully on high level. For example, in the healthcare sector, the main purpose of utilization of ML technique is to improve health outcomes, reduce healthcare costs, and advance clinical research. But, most hospitals are not presently taking decision to set up ML solution. One reason behind this is that the healthcare professionals are unaware to lack of the ML capability to make a successful system, install it in production, and put together with the clinical workflow. With the intention to formulate ML system is simple to use and to decrease insist for human experts. AutoML has come out as a rising field that look for to automatically choose, create, and parameterize ML models, as a result to achieve most advantageous outcome on a given task and/or dataset [32]. ✓✓ Deep Learning: Though DL has got remarkable achievement in many fields but it has to go away long journey. At a moment, many opportunities are there for enrichment. However, DL has highly developed world more rapidly

AI: An Imaginary World of Machine  179 than ever; there are many sectors still carry out to go. At present, we are too far from detailed knowledge of working of DL, how to get skillful machines, how machine becomes more elegant than the humans or can be able to learn similar to the human. DL has been resolving various problems at the same time and taking technologies to another level. Machine visualization, automatic text generation, self-directed vehicles, etc., are the several areas of AI where the DL is used at highest level. However, there are several complex problems for humankind to handle. For example, the people are currently expiring caused by hunger and food crisis, cancer and other dangerous diseases. We expect that DL and AI will be much more dedicated to the improvement of humankind, to perform the toughest scientific researches and to make the world a better place for every single human [33, 34].

9.8 Practical Implementation of AI Application In this section, we will discuss practical implementation of conversion of textual Gujarati joint characters which is in the form image into editable form using artificial NN. The proposed system takes an image as an input which contains textual Gujarati Joint-characters; initially, image is process using preprocessing methods of image processing technique after that it generates a feature vector of the joint character. In this proposed work, we have considered image file that contains 10 Gujarati joint-characters shown in Figure 9.4. For experimental purpose, we have prepared two different sizes of dataset such as train and test dataset of joint characters. We have taken 100 characters of each 10-Gujarati joint-characters given in Figure 9.4. Each character is processed using the proposed system and generates a feature vector of the character in the hand. These feature vectors are collectively known as train dataset. The system is trained using the well-known classification algorithm called a NN that assigned a label to the feature vectors. Now, when we apply a test image of joint characters to the propose model, the system would be processed the characters in the image and generate feature vector known as test pattern. The test pattern is evaluated with the patterns in the train dataset and would return associated label of the most matched pattern which will be finally convert into editable textual form. The complete operational algorithm of the proposed system is given as follows: (i) The input image is converted into Grayscale type. (ii) Grayscale image is converted into binary image so that the image contains only two values black and white. (iii) Using the in-built Sobel method, find the edges in intensity image. (iv) Create a square morphological structuring element of width w.

Figure 9.4  Sample Gujarati joint-characters.

180  The Smart Cyber Ecosystem for Sustainable Development (v) Dilate the binary image in text with square structuring element and then fill up the small hole in the image. (vi) Assign the label to each object (i.e., joint Gujarati character) present in the image and also counts number of object in the image. (vii) Locate the object in the image file and a boundary is created around each object for further processing. (viii) For each object of the image file, the features are extracted using feature extraction methods and prepared a vector known as feature vector. The feature vector of each character is placed in column by column manner and produced a feature matrix known as dataset. (ix) Generate a NN for train dataset. (x) Evaluate patterns (i.e., features) of train network with test pattern that would return associated label of most matched train pattern. (xi) Finally, recognized character is displayed in editable form. We have evaluated the performance of the proposed model for both similar as well as different font size of text image and would be achieved good accuracy rate. In both the cases, we obtained success rate nearer to 100% in most of the joint Gujarati character. A typical test character image having joint Gujarati characters of different font size are given in Figure 9.5. The proposed system is primarily normalized the character of different size into an ordinary one subsequently the further feature extraction process is performed. The size of the feature vector of each character is 35. There are 100 characters are used for each 10 characters given in Figure 9.4. Therefore, the numbers of such feature vectors are 1,000 and the size of train dataset is consisting of 1,000 trained patterns. We have trained the model using NN with the detailed information of train as well as test network of Gujarati characters, as shown in Figure 9.6. The outcome of system is nothing but it is a recognized character. The recognized Gujarati character is displayed in the editable form as shown in Figure 9.7. In Figure 9.7, the test image of different font size and the outcome of the system should be labeled as “Test image” and “Result”, respectively. The output of the system is shown as recognized characters in the form of editable textbox. This is the sample output which shows textual test image in editable form. Likewise, we can also take a test document in the form of textual image as input and convert it into editable text document too. Moreover, we can further extend the system for different font styles and color. We can also conclude that same process will be also repeated for handwritten Gujarati characters.

Figure 9.5  Test characters.

AI: An Imaginary World of Machine  181

Figure 9.6  Neural network training details.

Figure 9.7  Output window.

182  The Smart Cyber Ecosystem for Sustainable Development

References 1. Tim Jones, M., Artificial Intelligence Application Programming, pp. 3–4, Dreamtech Press (second Edition), 19-A, Anasari Road, Daryaganj, New Delhi-110002, 2006. 2. Jefferis, D., Artificial Intelligence: robotics and machine evolution, pp. 6–7 & 16–17, Silverdale Books, Bookmark Ltd, Desford Road, Enderby, Leicester, 2000. 3. Zhao, J., Liang, B., Chen, Q., The key technology toward the self-driving car. Int. J. Intell. Unmanned Syst., 6, 1, 2–20, 2018. 4. Wang, S., Generative Adversarial Networks (GAN): A Gentle Introduction, Tutorial on GAN in LIN395C: Research in Computational Linguistics; University of Texas at Austin: Austin, TX, USA, April 2017. 5. Pannu, A., Artificial Intelligence and its Application in Different Areas. Int. J. Eng. Innovative Technol., 4, 10, 79–84, April 2015. 6. Patterson, D.W., Introduction to Artificial Intelligence and Expert system, pp. 2–3, Prentice Hall of India (PHI), M-97, Connaught circus, New Delhi-110001, 1999. 7. Mosavi, A. and Varkonyi-Koczy, A.R., Integration of Machine Learning and Optimization for Robot Learning. Adv. Intell. Syst. Comput., 519, 349–355, 2017. 8. Vargas, R., Mosavi, A., Ruiz, L., Deep Learning: A Review. Adv. Intell. Syst. Comput., 5, 2, 1–12, 2017. 9. Hurwitz, J. and Kirsch, D., Machine Learning for Dummies, pp. 12–13, John Wiley & Sons, Inc, 111 River St. Hoboken, NJ 07030-5774, 2018. 10. Tecuci, G., Artificial Intelligence. WIREs Comput. Stat., 4, 2, 168–180, March/April 2012. 11. Song, X.E. and Xue, T., General Study about Applications of Artificial Intelligence in Physics. Int. J. Innovative Stud. Sci. Eng. Technol. (IJISSET), 5, 3, 43–48, 2019. 12. Chandra, Prof. K. and Rachana, S.H., Face Recognition in Facebook and Iphone X – A Review. Int. J. Recent Trends Eng. Res. (IJRTER), 04, 04, 165–170, April-2018. 13. Mainenti, D., User Perceptions of Apple’s Face ID. Inf. Sci. Hum. Comput. Interact., 1–16, 2017. 14. Ingle, S. and Phute, M., Tesla Autopilot: Semi Autonomous Driving, an Uptick for Future Autonomy. Int. Res. J. Eng. Technol. (IRJET), 03, 09, 369–372, Sep-2016. 15. Kalra, H.K. and Chadha, Dr. R., A Review Study on Humanoid Robot SOPHIA based on Artificial Intelligence. Int. J. Technol. Comput. (IJTC), 4, 3, 31–33, March 2018. 16. Mehta, H., Kanani, P., Land, P., Google Maps. Int. J. Comput. Appl., 178, 8, 41–46, May 2019. 17. Kepuska, V.Z. and Bohouta, G., Next generation of virtual personal assistants (Microsoft Cortana, Apple Siri, Amazon Alexa and Google Home). IEEE Conference, 2018. 18. Artificial Intelligence in Society, pp. 48–70, OECD publishing, Paris, https://doi.org/10.1787/ eedfee77-en, 2019. 19. Bhosale, S.S., Salunkhe, A.G., Sutar, S.S., Artificial Intelligence and its application in different areas. Int. J. Adv. Innovative Res., 7, 1(VI), 35–39, January-March, 2020. 20. Abduljabbar, R., Dia, H., Liyanage, S., Bagloee, S.A., Applications of Artificial Intelligence in Transport: An Overview. Sustainability, 11, 189, 1–24, 2019. 21. Lee, J., Kim, J., Kim, I., Han, K., Cyber Threat Detection Based on Artificial Neural Networks Using Event Profiles. IEEE Access, 7, 165607–165626, 2019. 22. Burrell, J., How the machine ‘thinks’: Understanding opacity in machine learning algorithms. Big Data Soc., 3, 1–12, January–June 2016. 23. Artificial Intelligence and Machine Learning: Policy Paper. Internet Soc., 11/10 Plaza America Drive, Suite 400, Reston, VA 20190, USA, 1–13, April 2017. 24. Harkut, D.G., Kasat, K., Harkut, V.D., Artificial Intelligence - Scope and Limitations, IntechOpen Limited, 5 Princes Gate Court, London, SW7 2QJ, UK, 2–3, 2019.

AI: An Imaginary World of Machine  183 25. Ghosh, A., Chakraborty, D., Law, A., Artificial Intelligence in Internet of Things. Inst. Eng. Technol. Res. J., 4, 208–218, 2018. 26. Artificial Intelligence and Machine Learning: Industry Insights and Applications (white paper), in: Infosys Limited, 2019. 27. Kumar, M., An Incorporation of Artificial Intelligence Capabilities in Cloud Computing. Int. J. Eng. Comput. Sci., 5, 11, 19070–19073, Nov. 2016. 28. Gaurav, S., Top Cloud and AI Trends for 2018, https://dzone.com/articles/top-cloud-and-aitrends-for-2018, 2018. 29. Kumari, S., Abhishek, R., Panda, B.S., Intelligent Computing Relating to Cloud Computing. Int. J. Mech. Eng. Comput. Appl. (IJMCA), 1, 1, 5–8, 2013. 30. Sheldon, A., Top 5 trends of Artificial Intelligence (AI), https://medium.com/hackernoon/ top-5-trends-of-artificial-intelligence-ai-2019-693f7a5a0f7b, 2019. 31. Tolba, A.S., El-Baz, A.H., El-Harby, A.A., Face Recognition: A Literature Review. Int. J. Signal Process., 2, 2, 88–103, 2006. 32. Waring, J., Lindvall, C., Umeton, R., Automated machine learning: Review of the state-of-theart and opportunities for healthcare. Artif. Intell. Med., 104 (Article 101822), 1–12, 2020. 33. Minar, M.R. and Naher, J., Recent Advances in Deep Learning: An Overview, 2018. 34. Ongsulee, P., Artificial Intelligence, Machine Learning and Deep Learning. Fifteenth International Conference on ICT and Knowledge Engineering, 2017.

10 Impact of Deep Learning Techniques in IoT M. Chandra Vadhana*, P. Shanthi Bala and Immanuel Zion Ramdinthara Department of Computer Science, School of Engineering and Technology, Pondicherry University, Puducherry, India

Abstract

Deep Learning (DL) has significantly changed the way process of computing devices human-centric content such as speech, image recognition, and natural language processing. It is very useful for safety-critical applications such as driverless cars, aerospace, defense, medical research, and industrial automation. DL models exceed human-level performance in terms of accuracy. It is a subset of machine learning that performs end-to-end learning and can learn unsupervised data and also provides a very versatile, learnable framework for representing visual and linguistic information. DL plays a major role in IoT related services. It serves as an emerging solution for developing IoT systems enhanced with efficient, reliable, and effective DL models. The amalgamation of DL to the IoT environment makes the complex sensing and recognition tasks easier. It helps to automatically identify patterns and detect anomalies that are generated by IoT devices. This chapter discusses the impact of DL in the IoT environment. Keywords:  Internet of Things (IoT), deep learning (DL), neural network, convolutional neural network (CNN), big data, cloud, recurrent neural network (RNN), RFID

10.1 Introduction Internet of Things (IoT) is a network of objects that are accessible via the Internet21. All these objects interact with either internally or externally. The challenge that exists in the environment during object communication and sensing is to make the decision properly and efficiently to control the environment. It is mentioned that the number of devices to be connected rise to 50 billion in 2020 according to a report by Cisco. Each IoT device generates an enormous amount of unstructured and structured data. So, there is a need for automatic data analysis to make the decision effectively. In IoT, the environment becomes smarter and makes transport cities more intelligent. IoT’s main objective is to allow the device to be connected to anything using a network or service at any time, any place. IoT provides better communication than machine-tomachine (M2M), Wireless Sensor Networks (WSNs), GSM, GPRS, microcontroller, GPS, microprocessor, etc. IoT is a fusion of hardware and software. It has the capacity and *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (185–214) © 2021 Scrivener Publishing LLC

185

186  The Smart Cyber Ecosystem for Sustainable Development versatility to adapt to the environment. These systems allow the user to achieve deeper automation, analysis, and integration within a system [1]. Technology makes the environment so modernized that improves the quality of life. Deep Learning (DL) helps to make such kind of decision. It is a kind of Neural Networks (NNs) which is a part of machine learning and uses an NN to simulate human-like decision-making. DL is used in the making of industrial robotics and processes an enormous amount of data when compared to machine learning and it is suitable for the IoT environment. DL takes a long time to train up the data and when it comes to hardware dependency and it requires GPU to train the data. This can be tuned in different ways but in machine learning; it has limited tuning capabilities. DL requires three layers comprises an input layer, an output layer, and a hidden layer. To take the input data, input layer is used and the hidden layer learns by itself so it was a good learning process and it tends to be more accurate and efficient for performing various computations on input data. These NNs are used to predict the output and perform classification on the data. DL requires a minimum time to infer information than any other method. IoT devices generate a huge volume of data and require real-time communication so the use of the DL model provides better results than other models is safe. It aims to improve learning efficiency, user experience, and network traffic by leveraging DL for IoT applications. All the DL models allows the storage of information. Enabling DL in IoT devices provides an efficient way to examine the unstructured data and act intelligently to both the user and the environment. It also provides a quality network connection in IoT devices. The integration of DL in IoT makes the system to capture and understand the environment easily and act accordingly. The next section provides detailed information about IoT characteristics, architecture, security requirements, challenges and applications, advantages, and disadvantages.

10.2 Internet of Things IoT is the most trending technology now and has a great impact on this modern society. Nowadays, many enterprises are moving to digital businesses and facilitating new business models to enhance the business by incorporating more employees and customers and leads to better efficiency. IoT makes people smarter and changing their day-to-day life by gaining control over their life by providing smart devices to automate homes, transportation, etc. It is also used for business and acts as a backbone for most of the important sectors like industry, healthcare, finance, retail, manufacturing, and agriculture. IoT sets a unique path in the technology world and it will be more useful for the stream as most of the businesses are now understand the potential of the connected device to keep them more competitive. In the IoT environment, the devices are communicated with each other and share their information through the Internet. Each device produces a huge volume of data and that can be collected and analyzed, and it is used to initiate an action, plan, manage, and decisionmaking. IoT consists of digital machines, mechanical and computing devices, objects, etc., and all are provided with a unique identifier (UID) and it can carry the data without any intervention in the network. IoT took the advantages from M2M communications and all the devices are connected to the cloud for managing the data. In the IoT scenario, all the sensor nodes, applications, and the smart devices produce and exchange the real-time data on real time. It senses the environment and acts according to that environment.

Impact of Deep Learning Techniques in IoT  187

10.2.1 Characteristics of IoT • Heterogeneity: All the devices in IoT are considered to be heterogeneous as it works on different platforms and networks. So, the interaction will take place with other devices through other networks. • Safety: It includes the protection of data. It is important to secure all the endpoints and the whole network. • Dynamic changes: All the connected devices can be changed constantly and the state of the device also changes dynamically. • Interconnectivity: In IoT, everything is interconnected with global communication and information. • Enormous scale: The management of data should be efficient and the communication with the devices should be managed. • Things-related services: IoT provides things-related services. To provide this service, the information world will also get changed and the physical world will also get changed accordingly. • Connectivity: This connectivity provides network accessibility and compatibility.

10.2.2 Architecture of IoT The architecture of IoT is composed of several layers by supporting different technologies [35]. • • • •

Smart device/sensor layer Gateways and networks Management service layer Application layer

10.2.2.1 Smart Device/Sensor Layer This is considered to be the lowest layer that consists of smart objects and it is integrated with sensors. These sensors are used to process and collect information from the real world. It contains various sensors that are used for various purposes. Mostly, sensors can measure temperature, speed, pressure, humidity, movement and electricity, air quality, etc. Figure 10.1 represents the architecture of IoT. A sensor is capable of measuring the property of the device and it converts into the signal. All the sensors are grouped according to their purpose. Mostly sensor requires connectivity to the sensor gateways. Sensors normally consume very low electricity and low data rate connectivity. With the use of WSNs, sensor nodes can be accommodated so that the sensor nodes can retain battery life and it covers larger geographic areas.

10.2.2.2 Gateways and Networks Even the tiny sensor will produce a massive amount of data and requiring robust and high performance. Current networks use various protocols and it can be adapted to use M2M networks. With the demand of these, it provides a range of services in IoT and it includes

188  The Smart Cyber Ecosystem for Sustainable Development Architecture of IoT APPLICATION LAYER

IoT APPLICATIONS (SMART CITY, SMART BUILDINGS, SMART HOMES)

MANAGEMENT SERVICE LAYER

MANAGEMENT CAPABILITIES, DATA MANAGEMENT, IoT BUSINESS PROCESS MANAGEMENT, ENTITY, ANALYTICS PLATFORM, SECURITY

GATEWAYS AND NETWORK LAYER

GATEWAY NETWORK, NETWORK CAPABILITIES, TRANSPORT CAPABILITIES

SENSOR LAYER

SENSOR NETWORKS (WI-FI, ZIGBEE, ETHERNET, BLUETOOTH) SENSORS (GPS, RFID, INFRA-RED, GYROSCOPE, ELECTRO-CHEMICAL)

Figure 10.1  Architecture of IoT.

applications such as context-aware applications, and high-speed transactions. In IoT, even though the devices are heterogeneous, all the protocols need to coordinate with each other to make the system work properly.

10.2.2.3 Management Service Layer It processes the information through the management of devices, process modeling, security controls, and analytics. This layer provides business and data management capabilities. The rule engine supports the creation of decision logic and triggers interaction to enable the IoT system more responsive. This layer uses data management to manage the data information flow. Particularly, in this layer, the information is accessed, integrated, and controlled. The security feature is implemented in all layers of IoT architecture. Security may be used to reduce the probability of risks.

10.2.2.4 Application Layer This layer includes many applications of IoT domains such as agriculture, healthcare, supply chain, culture and tourism, environment and energy, retail, and industries.

10.2.2.5 Interoperability of IoT In IoT, interoperability occurs in a different layer in the environment that exchanges the data and that needs to have communication. Figure 10.2 shows the dimensions of interoperability. The Interoperability then describes computer systems’ ability to exchange data and uses the information efficiently. Usually, the interoperability of IoT is complex. The main causes for this complexity are as follows. • Highly heterogeneous • Non-linear

Impact of Deep Learning Techniques in IoT  189 Dimensions of Interoperability

Organization Interoperability Semantic Interoperability Syntactical Interoperability Technical Interoperability

Figure 10.2  Dimensions of interoperability.

• Dynamic • Model available in many formats

10.2.2.5.1 Types of Interoperability • • • •

Technical interoperability Syntactical interoperability Semantic interoperability Organizational interoperability

10.2.2.5.1.1 Technical Interoperability

It addresses both hardware and software that enable M2M communication. It also deals with the transmission of bits. It ensures basic connectivity established between physical and logical connections. This type of operability is based on communication protocols and provide infrastructure to operate.

10.2.2.5.1.2 Syntactical Interoperability

It is associated with well-defined syntax and encoding for a particular service. High- level syntax such as HTML, XML, or ASN.1. In this, the message content needs to be serialized and sent in a proper format over the channel. The sender encodes the data using syntactical rules. Then, the receiver decodes the received message using the same syntactic rules. The problem arises when the encoding rules for the sender contradict the decoding rules for the receiver.

10.2.2.5.1.3 Semantic Interoperability

Semantic interoperability provides communication with various devices in IoT and helps to dynamically operate various resources that have different descriptions and can achieve better communication. It is associated with the concerns of the human rather than machine

190  The Smart Cyber Ecosystem for Sustainable Development interpretation. It is highly application-oriented. It is based on the understanding between people and the content that is exchanged.

10.2.2.5.1.4 Organizational Interoperability

This ensures all industries maintain the same pattern of structure in the organization and it makes it easier to manage multiple clients. It deals with the different organizations that can communicate and transfer the information using various information systems over different geographic regions. It mainly depends on the above three interoperability of IoT.

10.2.2.6 Security Requirements at a Different Layer of IoT The security requirements of IoT occur in different layers. These are as follows.

10.2.2.6.1 Application Layer

This layer requires authentication, application-specific encryption, cryptography, authorization, privacy, and policy management for security in IoT.

10.2.2.6.2 Service Support Layer

This layer requires secure computation, cloud computing, data aggregation, protected data management, handling, and cryptographic data storage.

10.2.2.6.3 Network Layer

This layer requires communication and connectivity security, cross-domain data security handling, and secure sensor/cloud interaction.

10.2.2.6.4 Smart Object/Sensor

This layer requires trust anchors and attestation, data format and structures, access control to nodes, and lightweight encryption.

10.2.2.7 Future Challenges for IoT There are various challenges in the IoT environment. Hence, there a need to address all the issues for achieving effective communication and automation. Figure 10.3 describes the future challenges of IoT. The main key challenges for IoT are privacy and security, cost and usability, data management, and energy preservation issues.

10.2.2.8 Privacy and Security In IoT, privacy and security are the key elements for the usage of IoT and it needs trust and security functions adequately. For this, it is necessary to provide trust and quality in the information and also sharing the information across applications. It is necessary to provide the exchange of information between IoT devices and user information securely. It also protects mechanisms for affected devices.

Impact of Deep Learning Techniques in IoT  191 Future Challenges for IoT FUTURE CHALLENGES OF IoT

PRIVACY AND SECURITY COST AND USABILITY DATA MANAGEMENT ENERGY PRESERVATION

Figure 10.3  Future challenges of IoT.

10.2.2.9 Cost and Usability IoT uses physical objects to connect through the internet and the cost of components is also growing that should support capabilities such as sensing and control mechanisms.

10.2.2.10 Data Management Data management is a crucial part of IoT. Considering the devices in IoT, all are interconnected and constantly exchanging information within the devices. It is also important that the amount of data generated and the mechanism involved in managing the data.

10.2.2.11 Energy Preservation It efficiently explains the interconnection of things and consuming less energy. It is a more important factor in IoT.

10.2.2.12 Applications of IoT IoT is used in several real-time applications such as agriculture, hospitality, traffic monitoring, water supply, energy, and the environment. The major applications include smart home, smart energy management, smart industry, smart cities, and smart farming.

10.2.2.12.1 Smart Living

Smart living requires remote control systems to turn any system on/off remotely and save electricity. The weather is used to view the weather conditions such as temperature, humidity, wind speed, and rain levels. Smart home appliances include a refrigerator LCD screen that tells the items inside and the expiry date of the ingredients and the ingredients to be purchased in the future, and all the information is accessible through an app. Washing machines make it possible to remotely track the laundry by changing the temperature control and controlling the self-cleaning feature. Figure 10.4 illustrates the IoT applications.

192  The Smart Cyber Ecosystem for Sustainable Development Applications of IoT SMART LIVING

SMART CITIES

SMART AGRICULTURE

SMART ENERGY

IoT

SMART INDUSTRY

SMART ENVIRONMENT

Figure 10.4  Applications of IoT.

Security monitoring is used to track cameras, monitor home security systems, and make people feel better. To avoid unauthorized entry, the intrusion detection system is used for opening the doors. To save costs and money, it is used for tracking energy and water usage at home.

10.2.2.12.2 Smart Cities

Smart cities include the features to track buildings and bridges’ material conditions. It also monitors any vibrations in the buildings to avoid loss. Lighting offers street lights with smart and weather-adaptive lighting. Security features include fire suppression, emergency alert systems, and wireless video surveillance. Transportation can help in any unexpected events by issuing proper warning messages and necessary diversions in the highways and smart roads. Smart parking is used to track the free parking spaces in the city and making the user find the closest available parking space. Waste management allows garbage level detections in the containers so that they can improve routes for the collection of wastages. With the Radio Frequency Identification (RFID) tags, the garbage cans are mounted which warn the worker to indicate to throw off the waste.

10.2.2.12.3 Smart Environment

Smart environment monitors CO2 emissions from factories and pollution generated by vehicles. Forest fire detection is used to track the gas and to establish warning zones. Weather monitoring comprises pressure, early earthquake detection, humidity, and temperature. Water quality regulates the quality of drinking water and the measurement of water suitability in rivers. During rainy periods, river floods are used to control the water level in the dam and reservoirs.

10.2.2.12.4 Smart Industry

Smart industries include the features to identify explosives and poisonous gases, level of gas leakage and monitor hazardous gas and oxygen levels in chemical plants and also monitor oil and gas levels in storage tanks and pipelines. Maintenance and repair provide early

Impact of Deep Learning Techniques in IoT  193 identification of any malfunctioning units in a smart industry and maintenance can be carried out automatically for some failure.

10.2.2.12.5 Smart Energy

Smart energy requires a smart grid that monitors industries’ energy use. Wind powerhouse helps to monitor the flow of energy from wind turbines and power plants. Smart meters are used for the analysis of consumption patterns. Power supply controllers include AC-DC power supply controllers linked to consumer electronics and telecommunications applications. Photovoltaic installations are used to track the efficiency of solar power plants and optimize them.

10.2.2.12.6 Smart Agriculture

Smart agriculture helps to enhance the food production. To prevent microbial and other contaminants, compost is used to control humidity levels and temperature levels. Animal monitoring is used to locate and classify animals that graze in open fields, and air quality and ventilation on farms are also studied. In animal farms, offspring treatment tracks the growth conditions of the offspring and ensures the sustainability and well-being of the plant. It reduces spoilage and wastage from plants through regular control of accurate data and field management.

10.2.2.13 Essential IoT Technologies The most important five technology used to develop products and services based on IoT. These are as follows. 1. 2. 3. 4. 5.

Radio Frequency Identification Wireless Sensor Networks Middleware Cloud Computing IoT Application Software

10.2.2.13.1 Radio Frequency Identification

RFID allows the process of identification and captures the data automatically using radio waves, tags, and readers. Barcodes store the minimum amount of data but tag stores more data [26]. This includes data in the Electronic Product Code (EPC) format; it is considered as global RFID. Three tag types, namely, Passive RFID, Active RFID, and semi-passive RFID, are used. The Passive RFID tags are mainly based on a radio frequency that is transmitted from the reader to the tag and it is not powered by the battery. Tags are mainly used in passports, digital rolls, item-level monitoring, and supply chains. There is a battery supply for active RFID tags and it can have communication with the reader. It also consists of external sensors for controlling temperature, pressure, and other conditions and is also used in manufacturing, laboratories, and hospitals. Using batteries, semi-passive RFID tags power up from the reader. There is a very high cost for active and semi-passive tags.

194  The Smart Cyber Ecosystem for Sustainable Development

ESSENTIAL IoT TECHNOLOGIES

Essential IoT Technologies

RADIO FREQUENCY IDENTIFICATION(RFID) WIRELESS SENSOR NETWORKS(WSN) MIDDLEWARE CLOUD COMPUTING IoT APPLICATION SOFTWARE

Figure 10.5  Essential IoT technologies.

10.2.2.13.2 Wireless Sensor Networks

It mainly consists of autonomous sensor systems for tracking the external surroundings with the help of RFID systems. These are used for keeping track of current temperature and position. Figure 10.5 shows the essential IoT technologies. It also allows for different networks and multi-hop communications. With the use of wireless communications, it provides efficient, low cost, low power communication. It is also used for maintenance and tracking system for vehicles and also used in cold chain logistics.

10.2.2.13.3 Middleware

It acts as a software layer that is accessed between applications to perform communication, input, and output. It hides the information of various technologies from other services and that it is not related to IoT applications. It also paves the way for the development of new applications and services and which is important for IoT application development. It also plays a significant role in integrating existing technologies into new ones. Most of the middleware architecture for IoT follows a service-oriented approach to aid the network topology.

10.2.2.13.4 Cloud Computing

Cloud Computing is an architecture for accessing the shared resources in the network. The resources can be networks, software, applications, and services. It needs a large storage space, fast decision-making processing speeds, and high-speed broadband access to stream audio or video. So, it allows a back-end solution for storing huge real-time data in IoT and humans.

10.2.2.13.5 IoT Application Software

All the application software is used for the development of an organization. It provides physical connectivity between devices and networks but IoT provides interaction between humans and devices effectively. These IoT applications are used to ensure message reachability and provide reliable communication between the end-users.

Impact of Deep Learning Techniques in IoT  195

10.2.2.14 Enriching the Customer Value The usability of IoT devices is increasing gradually day by day. By utilizing the IoT devices, the organization leads to the effective implementation of IoT applications. Usability is based on the categorization of IoT for enterprises, growing technology, and literature review of IoT, and it is classified into three types: 1. Monitoring and control 2. Big data and business analytics 3. Information sharing and collaboration

10.2.2.14.1 Monitoring and Control

It allows the user to collect data on performance, usages, and environmental conditions, and it allows managers to keep track of performance anytime and anywhere. Most recent control and monitoring systems such as smart meters and smart grid are used to forecast outcomes and automate operations, identify the areas where modifications are required, and expose operational trends that result in lower costs and high productivity. A smart home is considered as one of the most monitoring and control systems in the IoT environment. Its principal work is to provide protection, continuous monitoring, and saving energy. It can be controlled from outside to a particular home through a computer or smartphone or tablet. For this monitoring and control system, Verizon Home Monitoring22 and control network is one of the best solutions. It was designed specifically for remotely controlling home automation applications and it uses wireless communication technology. This network allows the user to control and adjust the lights and temperature, to manage the security system, to lock and unlock the doors automatically, and able to receive automatic notifications for a particular event. It is also used in monitoring and controlling various components of the car also. Ford and Intel are joined together for the new project and exploring some more chances to customize the user experience by giving comfort. So, the customer values are properly indexed and integrated into another new environment to provide a profit to Ford.

10.2.2.14.2 Big Data and Business Analytics

In upcoming years, business analytics tools such as wearable health monitoring sensors can be built into IoT devices which helps to decide according to the status. In IoT, all the devices are equipped with sensors and produce a significant amount of data and this transforms business intelligence and decision-making tools for clients. All the data is used to solve business problems and improve user satisfaction. With the help of business analytics, it is possible to analyze huge data, for example, healthcare data. IoT uses healthcare services to take care of patients. Nowadays, IoT facilitates collecting and maintaining the patient data including their behavior and health regularly. It helps to take care of the patients carefully from remote places. For example, Humana’s health sense neighbor is one of the remote monitoring systems and it sends the behavioral pattern to human care managers through sensors and it also monitors daily routine with the help of data analytics and helps to prevent any unconditional events.

196  The Smart Cyber Ecosystem for Sustainable Development It also deals with IoT-based big data which is used to transform the healthcare industry. For example, the development of Oral-B Pro 5000 is considered as an interactive electric toothbrush that provides the user to have more personalized oral care. This also records the brushing habits of the user by using mobile technology and also giving tips for oral care. Using this electric toothbrush, brushing time is increased to 60 seconds when compared to manual one (i.e.) nearly 2 minutes and 16 seconds.

10.2.2.14.3 Information Sharing and Collaboration

Information can be shared between people and things. Usually, sensing is the first part of information sharing and collaboration. It helps to avoid information delay and distortion. To amplify the information sharing and collaboration, it is recommended to use Bluetooth Low Energy23 (BLE) and it is based on location-based technology in mobile which uses ultrasound to enhance the performance. This type of enhancement allows enhancing customer satisfaction and market revenue. The sharing of information leads to higher customer satisfaction.

10.2.2.15 Evolution of the Foundational IoT Technologies IoT technology consists of software, algorithm, network, data processing, and hardware. The network plays a significant role in IoT and it uniquely identifies the object [10]. In that, network technology is upgrading to wireless communication which allows applications to become more flexible. Objects make effective communication to enhance functionality and effectiveness with each other. The software should be developed with interoperability of IoT, security requirements, and privacy [24]. To provide information about the location of the devices, the data generated by IoT is to be aggregated and analyzed. It does not work well with traditional data processing during real time. All available data on the devices cannot be processed at all. Context-aware data processing uses particular information such as location and temperature. This enables certain devices to decide which data is to be collected and interpreted to provide relevant information to the user. This data processing will give useful information by identifying the current location of the user. Cognitive data processing requires human cognition that is changed into IoT applications. It is trained using algorithms based on AI. It uses image recognition techniques to identify the environment and then eventually processes the data it receives user feedback. It can predict, sense, and learn tasks and environments using AI algorithms. Optimization of data processing is the most challenging and time-consuming process because of the enormous amount of data produced in IoT devices. Big data applications help to optimize data processing such as environmental monitoring, smart manufacturing, and smart grids.

10.2.2.16 Technical Challenges in the IoT Environment The explosion of data generated by IoT devices and leads to many challenges in terms of security, user privacy, storage, server technology, and data center networking. There are also some other technical challenges such as data mining challenges, privacy challenges, and management challenges.

Impact of Deep Learning Techniques in IoT  197

10.2.2.16.1 Data Management Challenge

It is essential to store and process generated data by IoT devices. Organizations must invest in data storage and the data obtained from networks must be processed. Proper backup is based on values and needs. To increase the device response time and performance, the data center is widely distributed. Additional bandwidth is used primarily by IoT devices.

10.2.2.16.2 Data Mining Challenge

It is efficient to use a data mining tool for processing a large amount of data. The streaming data consists of location, temperature, humidity, movement, etc. All the data are trained using mathematical models. It is not suitable for unstructured data like audio and video. Then, it also requires more data analysts and advanced data mining tools that are used to stream the data from the networks.

10.2.2.16.3 Privacy Challenge

Protecting user information is a major task in IoT which is used to improve people’s quality and thus reduce the implementation cost. According to a survey, trust in IoT is only the benefit of smart devices with privacy concerns that were agreed by 22% of Internet users. IoT gains through the smart home system and wearables with confidence and acceptance of IoT which mainly depends on privacy protection.

10.2.2.17 Security Challenge When the IoT devices are connected over the internet, there is a potential threat. When a greater number of devices increased, there is also an increase in the potential attack by intruders and other cybercriminals. According to a survey, 70% of IoT devices have more vulnerabilities. Most of the devices have vulnerabilities due to insecure web interfaces, insufficient authorization, and inadequate software protection. IoT devices do not use any techniques to encrypt the data. Some of the IoT applications provide infrastructure and services. Lack of privacy and security will lead to severe problems and these security challenges were resolved by trained developers to embed security solutions that include firewall and IDS. These products encourage the users to make use of the security features of IoT.

10.2.2.18 Chaos Challenge The era of IoT technologies is a useful one but still, it has some issues like standards, privacy, insufficient security, complex communications, and not having proper devices. If the devices are not properly designed and the collaborative applications bring our lives into chaos. If one small error occurs in a network, then it brings the whole system down. If one sensor of the control system is not working, then the controller receives an error signal which may cause severe problems. A single device can cause many problems but when comes to the whole network, it affects all the devices that are connected to the network. To avoid confusion in this connected world, every company needs to reduce the sophistication of the connected devices, to improve the security and application standardization, and to ensure the user’s privacy and safety of the user at their level of satisfaction.

198  The Smart Cyber Ecosystem for Sustainable Development

10.2.2.19 Advantages of IoT 1. 2. 3. 4. 5.

Accessing information is easy from anywhere at any time. Transferring data over the network saves time and cost. Automated tasks help improving business and reduce human intervention. It provides efficient resource utilization. It provides enhanced data collection.

10.2.2.20 Disadvantages of IoT 1. It has design complexity issues in maintaining IoT technology. 2. Security is the main issue when the devices are connected to the public network. 3. User information is not preserved in the network. 4. There is no international standard compatibility for IoT as the devices are manufactured by different vendors. 5. If a greater number of devices gets increased, collecting and storing data is more tedious. IoT devices are mainly used for processing the data in the environment and it produces a huge volume of data continuously. For handling those volumes of data, DL provides an efficient way of handling complex data in the IoT environment. The upcoming section briefly explains DL techniques and applications.

10.3 Deep Learning DL deals with big data for scalability and generalization. It is classified as three types of learning such as supervised, semi-supervised, or unsupervised. It is used for computer vision, image and audio processing, and pattern recognition and is also used for the improvement of classification and prediction problems [3]. All the complex deep neural algorithms are trained using powerful GPUs. Supervised learning consists of three types of datasets. The first one consists of a training set and it is considered as a learning process [32]. This enables learning algorithms to be supervised and it also contains the exact results under the labeled feature. Internal layers of NNs are calculated by their weights and are determined by data and their expected results. Finally, optimal weights are obtained. The second type is the validation set that assists the learning process in tuning the optimum weight parameters. The third type is the testing set that is used for evaluating the efficiency of the training process. Before the commencement of the learning process, all the datasets are split into two sets such as the training set and testing set. The validation dataset is obtained from the training set. In the NN algorithm, epochs are considered as learning time. When considering one epoch the training set is passed through the network completely. Every NN training algorithm aims to determine the best possible weight values for the problem [25]. For determining the optimal set of weight, a trade-off is required between minimizing the network error, computational time, and

Impact of Deep Learning Techniques in IoT  199 maintaining the network. Neurons are the main part of the learning process and it takes one or more input from previous neurons with weights and substitute in the activation function and finally produces the output.

10.3.1 Models of Deep Learning This section briefly explains various DL models, their attributes, characteristics, and applications [30]. These are as follows. 1. 2. 3. 4. 5. 6. 7. 8. 9.

Convolutional Neural Networks Recurrent Neural Networks Long Short-Term Memory Autoencoders Variational Autoencoders Generative Adversarial Networks Restricted Boltzmann Machine Deep Belief Network Ladder Networks

10.3.1.1 Convolutional Neural Network Deep Neural Network provides the dense connection between the layers and it is very difficult to train [31]. The main reason is that such models have translation inversion properties. In an image, it does not learn the features. But by endorsing translation equivariance computations, CNN solved this problem. In which, CNN receives 2-D input and extracts high-level functionality through the sequence of hidden layers. The convolutional layer is situated at the core and it has a set of parameters known as filters. In the training phase, each convolutional layer filter passes via the input volume and measures the input and filter. Then, entire input leads to the feature map filter. It also consists of pooling layers and it operates on feature maps. The pooling layer’s purpose is to minimize the spatial size by limiting the usage of parameters, computations and minimize the overfitting. The important function of CNN is the ReLU function (Rectified Linear Unit) with an activation function and neurons. It helps to complete the training task very faster than other models. CNN model comes under the category of discriminative and it is classified under supervised learning. Normally, CNN receives input in 2-D (image, sound, etc.). The main characteristics of CNN are convolutional layer plays a major role in computations and it provides less connection as compared to DNN. To incorporate CNN in IoT, applications require large training datasets.

10.3.1.2 Recurrent Neural Networks RNN is mainly used for prediction based on the previous samples and it also classifies the samples and analyzes the sequential inputs. The feed-forward network is not suitable for RNN because there is no possession between input and output layers [34]. The RNN input

200  The Smart Cyber Ecosystem for Sustainable Development consists of all the samples of that layer and it holds the previous sample and current sample. Always the RNN output concerning time is t-1 and the output time t is affected. Each neuron contains a feedback loop that returns the current output as input for the subsequent process. It has some internal memory and keeps recording information from the previous layer. For training the network, it is necessary to use Back Propagation Through Time (BPTT). RNN includes many hidden layers and it provides a memory instead of hierarchical representations. RNN is deeper and it includes more layers between input and hidden layers and also includes more layers between output layers and the hidden layer. RNN comes under the discriminative category and it follows supervised learning. RNN takes input in the form of sequence and time-series. RNN processes the data in a sequence manner via internal memory and offers an advantage for IoT. For IoT applications, it is used to identify movement patterns and behavior detection.

10.3.1.3 Long Short-Term Memory Long Short-Term Memory (LSTM) is based on the values of the gate and it computes values between 0 and 1 corresponding to the input. It is the continuation of RNN most of the functions follow the same pattern [32]. It also consists of neurons that are known as memory cells. It has read gate, multiplicative forget gate and write gate. It also has a feedback loop which is used to store information. These gates are used to give memory cell access control and to prevent irrelevant inputs. When the forget gate is turned off by sending 0, then the neuron also overlooks the recent content in the cell. The other neurons can write to those neurons when the write gate is assigned to 1. And also, if the read gate is assigned to 1, it can read the content of the neuron. The major change in RNN and LSTM is that in LSTM, the forget gates can control the cell states and also ensure that it should not affect the performance. The activation function for these gates is sigmoid or tanh. It causes a vanishing gradient problem during the training phase of some other models. LSTM stores computations in memory cells that should not be twisted over time. LSTM model’s performance is better than RNN models. This comes under the category of discriminative. It follows supervised learning methods. The input data for LSTM consists of long-time dependent data, time-series, and serial [18]. LSTM provides long-time log data with better performance. Accessing memory cells is highly protected by logic gates. The main IoT applications using LSTM are human activity recognition and mobility prediction.

10.3.1.4 Autoencoders AEs are made up of two layers such as input layer and the output layer. All the layers are connected through two or more hidden layers. It is also a form of the NN that is used to solve unsupervised learning problems and transfer learning [6]. It reconstructs the network by input and transfers all inputs into outputs in a simple manner. It is mainly used for fault detection and diagnosis tasks. With the help of these techniques, it can be applied for IoT applications. The main two components of AEs are encoder and decoder. It minimizes the reconstruction error.

Impact of Deep Learning Techniques in IoT  201

10.3.1.5 Variational Autoencoders Variational autoencoders (VAEs) are based on semi-supervised learning. It is suitable for IoT. It is also used for failure detection in sensing and IDS security systems. It consists of two network models one is used for generating samples and the others for preliminary inference.

10.3.1.6 Generative Adversarial Networks Generative Adversarial Network (GANs) are composed of two kinds of networks such as generative and discriminative networks. Both networks are working together to provide high-quality data and synthetic data. The generative network is designed to produce input data that misleads the discriminator [15]. The former network is responsible for generating new data then learns the distribution data from the training set. Then, the latter network also performs discrimination between fake and real data. The GAN’s main objective is to focus on minimax values where a network seeks to minimize the value and while another network maximizes the value. In the IoT scenario, GAT is used for creating something new on the available data. It also provides service for visually impaired peoples for text-to-speech conversions and image-to-sound convertors. GAN comes under the hybrid category and it follows semi-supervised learning. It uses various input data. It is suitable for noisy data. The IoT applications in GAN includes image-to-text, localization, and wayfinding.

10.3.1.7 Restricted Boltzmann Machine Restricted Boltzmann Machine (RBM) is considered as a stochastic ANN and is comprises of two layers such as the hidden layer and visible layer. Then, the hidden layer contains latent variables but the visible layer contains the input. RBM is mainly used for the connectivity of neurons. RBM should have a bipartite graph in which every visible neuron must link to all hidden neurons. There is no relation in the same layer between the two devices. This uses a backpropagation and gradient descent algorithm for training and optimizes the weights of the network. RBM features are identical to AE. It performs feature extractions for input data. RBM comes under the category of generative and it follows both supervised and unsupervised learning and it can perform tasks on various input data. It has an expensive training procedure and it is suitable for feature extraction, classification, dimensionality reduction. The applications of IoT in RBM is energy consumption prediction and Indoor localization.

10.3.1.8 Deep Belief Network Deep Belief Network (DBN) also consists of a visible layer and several hidden layers. It extracts the data from the training data in a hierarchical representation and forms the input data. Finally, it adds the SoftMax layer and it is used for prediction [7]. The training of DBN is performed layer by layer and the RBM training is done on the top of the layer. For this reason, compared to all the architecture of DL, this makes the DBN more powerful. It is considered a fast algorithm in DL.

202  The Smart Cyber Ecosystem for Sustainable Development Some of the applications are fault detection classification in industries and emotional feature extraction in images and threat identification in alert systems.

10.3.1.9 Ladder Networks Ladder Networks are used for several tasks which include image recognition and handwritten digits recognition. There are two encoders and one decoder in the architecture. The encoder also acts as a supervised part of the network. One encoder also called the clean encoder and performs computations and another encoder is known as the corrupted encoder and it includes Gaussian noise to all the layers [13]. Then, the decoder carried out unsupervised learning. The architecture’s main goal is to reduce the cost of supervised and unsupervised networks. It comes under the category of hybrid. It follows semi-supervised learning and it works on various input data. Ladder networks are suitable for noisy data. The applications of IoT in ladder networks are authentication and face recognition.

10.3.2 Applications of Deep Learning DL is used in many real-time scenarios [20]. Some of the applications of DL are as follows.

10.3.2.1 Industrial Robotics It empowers robots in industries to solve complex problems and to recognize real-world problems and provides a better solution than humans. It uses CNN and LSTM models to solve the problems in industries using robots.

10.3.2.2 E-Commerce Industries In e-commerce, it can store an immense amount of data and can analyze the online activity of millions of customers around the world using artificial intelligence and DL.

10.3.2.3 Self-Driving Cars Many vehicles are now using the driver-assist system to automate the driving experience. With the help of DL and artificial intelligence, it can do better in self-driving cars and it continues to improve safety and performance.

10.3.2.4 Voice-Activated Assistants Voice assistants run on DL algorithms which help to train themselves. All the voice assistants are trained using artificial intelligence models to understand human speech in different situations.

10.3.2.5 Automatic Machine Translation DL uses neural machine translation. It is a good choice for accurate translation and delivers better performance. It uses DNN to improve the system to make it more efficient.

Impact of Deep Learning Techniques in IoT  203

10.3.2.6 Automatic Handwriting Translation With the use of DL, the main goal is to achieve the highest accuracy with the use of CNN and LSTM models, it recognizes the characters and reconstructs them according to classification and segmentation.

10.3.2.7 Predicting Earthquakes It uses NN techniques and artificial intelligence to predict earthquakes. Some of the researchers are using machine learning techniques for predictions.

10.3.2.8 Object Classification in Photographs DL uses ImageNet for image classifications. It creates a model that accurately classifies the given images in the particular datasets. It uses CNN for image classification tasks in DL.

10.3.2.9 Automatic Game Playing It uses some learning techniques like supervised learning, RNN, and CNN with AI for automatic game playing.

10.3.2.10 Adding Sound to Silent Movies DL automatically recognizes the pattern of the movie. It trains the system to pair the right sound to the movie and it automatically adds sounds effects.

10.3.3 Advantages of Deep Learning 1. 2. 3. 4.

Automatically features are identified and tuned for desired features. Robustness in the data is automatically learned. Parallel computations can be performed with GPUs and it is scalable. Deep learning is more flexible to adapt to new problems.

10.3.4 Disadvantages of Deep Learning 1. 2. 3. 4.

The DL model is not suitable for small datasets. It takes a longer time to execute the model. It requires expensive GPUs to train the model and thus increase costs. It requires skilled persons to train the data, parameters and requires knowledge about selecting DL tools.

10.3.5 Deployment of Deep Learning in IoT In many research areas such as natural language processing, bioinformatics, and image recognition, it can handle enormous data that is continuously produced from IoT devices. DL

204  The Smart Cyber Ecosystem for Sustainable Development is more effective in managing such an immense amount of data. DL has been introduced into IoT and some mobile applications for the early predictions of data [28]. DL has high efficiency in handling complex data and it plays a major part in IoT.

10.3.6 Deep Learning Applications in IoT Deploying DL applications in IoT is more challenging as DL is used in several areas for analyzing results. Some IoT related applications use image classification, plant, or human disease prediction. Many IoT applications are still providing the best services. The services provided by the IoT environment are used in routine work which is represented as foundational services. The foundational services adapt the DL technology for the improved lifestyle. Some of the foundational services are as follows: 1. 2. 3. 4. 5.

Image Recognition Speech/Voice Recognition Indoor Localization Physiological and Physiological detection Security and Privacy

10.3.6.1 Image Recognition All the inputs in DL are in the form of image, audio, or video. The images are in high resolution that can be sent from one device to another. The most important task in DL is object detection and it is done by recognizing images. It is further processed for image classification and recognition. For DL, commonly used image datasets are CIFAR-10, VGG datasets, CIFAR-100, tiny image datasets, etc. These kinds of image datasets are very useful for image classification in DL techniques.

10.3.6.2 Speech/Voice Recognition Speech/voice recognition is used to interact with people through mobile phones and to interact with their devices. Nowadays, many peoples are started moving to automatic speech recognition and the possibility of having input interaction with the devices are also get reduced. The main concern of DL is to have automatic recognition on the devices. In the NN, voice is given as an input to the network, and the information is done through network hidden layers and eventually, the output is converted into speech. It consumes low energy and detects the sound by monitoring the environment.

10.3.6.3 Indoor Localization Indoor Localization describes the indoor navigation and location in the environment by providing location services and location marketing for merchants. The input data is generated from various applications using different technologies includes RFID, Bluetooth, Wi-Fi, and Infrared. Mostly, the approaches are based on Wi-Fi or Bluetooth which uses smartphone phones to receive transmitter signals. DL has been successfully implemented in indoor positions [36]. DL is made up of online localization phases and offline training.

Impact of Deep Learning Techniques in IoT  205 With the use of the CNN method, all the phases are based on the high accuracy of the number of hidden layers.

10.3.6.4 Physiological and Psychological Detection The combination of IoT and DL techniques are used to detect human activities such as emotions, pose, activity. Many of the IoT applications are used to analyze the human pose to deliver their outcome such as smart cars, Xbox, education, entertainment, sports, and manufacturing. The camera is more useful for capturing human activities and it transfers the video to DNN to find out the human pose. Beyond human pose, one can estimate the human emotion using CNN, AE, and DBN. It is a combination of two models such as layers of convolutional and RNN.

10.3.6.5 Security and Privacy The two important aspects of IoT are security and privacy. The key feature of the systems depends on securing the devices from intruders. The most common type of attack on the system is False Data Injection (FDI) as mentioned by [17]. The mandatory factors for the IoT are security and preserving privacy. Optimization of stochastic gradient descent is used in many approaches of DL.

10.3.7 Deep Learning Techniques on IoT Devices Most of the DL research focuses on IoT devices to develop models and algorithms for efficient operation. When DL is implemented in IoT [5], it requires power, memory, and processors, and resource-constrained devices for training the generated data using DL [12]. DL consist of a greater number of parameters and also in need of large storage resources for computing. Some technologies have been used for exploring DL techniques in IoT devices. These are as follows. 1. 2. 3. 4.

Network Compression Approximate Computing Accelerators Tiny Mote with DL

10.3.7.1 Network Compression Using network compression, the use of DNN for resource-constrained devices is more effective and the dense network is transformed into a sparse network. It helps to reduce the computational cost and storage of DNN [16]. It is only suitable for certain networks. AlexNet is also used in network compression. Especially, feed-forward layers are smarter than the convolutional layer. The activation function in DNN contributes time efficiency and also use a ReLU function than a tanh function. This method is based on unnecessary connections and also weight sharing mechanisms. It is calculated as each weight is having an n bit index that has 2n values. It is also based on pruning networks with many parameters.

206  The Smart Cyber Ecosystem for Sustainable Development

10.3.7.2 Approximate Computing The computing method is mainly relied on the implementation of machine learning tools on IoT to save energy to extend the lifetime of the devices. Implementation of DL models on resource-constrained devices is more effective and the NN usage is also acceptable. Using backpropagation, it is easy to identify the neurons with less accuracy. The approximate NN is developed by substituting fewer neurons in their original network. This technique is known as precision scaling. After network formation, the inputs and weights are used between accuracy and energy.

10.3.7.3 Accelerators The accelerators are used in IoT devices to improve the energy efficiency and memory of DL models [11]. The training of the method is very complex and energy-intensive. Some of the approaches are used to improve the smartness of devices are particularly designed for DNN and CMOS technologies. It has two algorithms at run-time, one is used for compressing the layers and another is used for decompressing the models on the processors.

10.3.7.4 Tiny Motes Tiny motes deal with small processors with strong DL models. The outcome of IoT applications is a time-consuming process and it makes the system to wait for their response [9]. It is mainly based on security and privacy for all the applications. It is efficient to deploy DL models for some limited data.

10.4 IoT Challenges on Deep Learning and Future Directions Some of the research challenges of IoT on deep learning models are 1. 2. 3. 4.

Lack of IoT dataset Pre-processing Challenges of 6v’s Deep learning limitations

10.4.1 Lack of IoT Dataset IoT consists of real-time datasets that lead to a lack of availability and the main cause for implementing DL models into IoT is that DL requires enormous data for processing to achieve more accuracy. Privacy and accessing copyright datasets are more important in datasets like education and healthcare. More numbers of the dataset are required for DL is more challenging.

Impact of Deep Learning Techniques in IoT  207

10.4.2 Pre-Processing The most important part of DL is pre-processing. Most of the DL requires processed data than raw data to yield good performance. In IoT, pre-processing is one of the tedious processes with different data sources.

10.4.3 Challenges of 6V’s In comparison with big data in DL, The main IoT characteristics and big data characteristics such as volume, velocity, variety, veracity, validation of data, and variability. All these big data characteristics pose a challenge for DL techniques. In the big data, it contains a huge volume of data that gives a challenge for DL and the volume of data for input, attributes, classification of DL models, and performance [4]. In big data, variety refers to the heterogeneous nature of data both structured and unstructured. IoT deals with a variety of data is also a challenging task for DL. Here, velocity defines the speed generation of data. Veracity refers to data accuracy, precision, and trust. The IoT big data of veracity also plays an important part in DL techniques. It also provides data analytics to improve the business in IoT.

10.4.4 Deep Learning Limitations The limitations are based on the prediction of images that are unrecognizable by a human being. Other limitations are based on DL models and many IoT applications need regression for their analytics. It provides a combination of DBN and supports vector regression (SVR) for regression analysis. Most of the predictions require DL with regression methods.

10.5 Future Directions of Deep Learning The future directions of DL include the following: 1. 2. 3. 4. 5. 6. 7.

IoT mobile data Integrating contextual information Online resource provisioning for IoT analytics Semi-supervised analytic frameworks Dependable and reliable IoT analytics Self- organizing communication networks Emerging IoT applications

10.5.1 IoT Mobile Data All IoT data is provided by mobile devices. For DL models in IoT, it offers an efficient way to use big data. It uses all the capabilities of big data analytics in DL and executes a MapReduce framework for IoT applications.

208  The Smart Cyber Ecosystem for Sustainable Development

10.5.2 Integrating Contextual Information The sensor understood the IoT environment and it provides complete information from different sources of data [33]. This helps to provide data analytics for that environment. In all situations, it uses contextual information and this helps the system to provide better action and react according to the environment.

10.5.3 Online Resource Provisioning for IoT Analytics It includes the fog and cloud implementation of DL-based data analytics. Online resource provisioning is required for data streaming in fog and cloud services. Analyzing the volume of data is not enough and it is necessary to use the algorithms for analyzing the stream of data [8]. Some of the authors proposed the algorithms such as DL mechanism and online auctioning algorithm to support online provisioning resources for cloud and fog and these IoT applications.

10.5.4 Semi-Supervised Analytic Framework To train a large amount of data, all the algorithms in DL comes under supervised learning. The machine learning algorithms used in the design of semi-supervised learning is used for applications like smart cities [29]. In smart cities, it requires a small training set and the use of an immense amount of unlabelled data also helps to provide accuracy.

10.5.5 Dependable and Reliable IoT Analytics It is mostly relying on the cyber-physical system (CPS) and IoT, as there is a need for security mechanisms to prevent the system from failures as well as attacks, which becomes more difficult. It is more efficient to have DL approaches for CPS to trace a large data in IoT devices and to identify the failure and vulnerability of the system. This helps to prevent the system from errors and recovery to improve the dependency level of devices like IoT and CPS.

10.5.6 Self-Organizing Communication Networks In the IoT devices, it is very hard to configure and maintain communications and networks. Whereas, DL provides self-organization with a set of services that includes self-healing, self-load balancing, and self-configuration as compared to traditional approaches.

10.5.7 Emerging IoT Applications Some of the emerging IoT applications are involved in future directions of DL are as follows: 1. Unmanned Aerial Vehicles 2. Virtual/Augmented Reality 3. Mobile Robotics

Impact of Deep Learning Techniques in IoT  209

10.5.7.1 Unmanned Aerial Vehicles It is considered the most promising applications that improve the delivery service in critical situations. Unmanned Aerial Vehicle (UAV) is used in many analyses such as search and rescue operations, inspections, and surveillance tasks. DL provides the best prediction and decision-making for UAVs [27]. It can be temporarily used for providing fog services and analytics.

10.5.7.2 Virtual/Augmented Reality The advantages are exploited by both IoT and DL [2]. It provides a service for object recognition and classification. It is mainly used for education and connected cars.

10.5.7.3 Mobile Robotics Robotics mainly used in many industries for setting and moving up the materials in the industry and for performing some tasks in hazardous environments [14]. It also provides several kinds of sensors such as cameras and LIDARs. It is also used to check the performance of CNN in vision-based tasks. In robotics, the most important thing is that DL models must able to respond to robots in real time.

10.6 Common Datasets for Deep Learning in IoT Datasets usually contain images or videos for image or object recognition, image classification, and prediction using images for DL [19]. The datasets for DL in IoT are indicated in Table 10.1. Some of the collections of common datasets5 suitable to use for DL in IoT are as follows.

10.7 Discussion In IoT, all the devices are connected with the network of physical objects or things. It can connect with many emerging technologies like DL, machine learning, and cloud. IoT is used to interact with other devices in the network for communication and it makes the devices always connected to the world. IoT allows to access the device remotely across the network infrastructure and able to sense the environment and it is used to connect all the devices. It collects, stores, and analyzes the data in the environment. During processing, IoT devices generate an enormous quantity of data. The vast amount of data in IoT is very difficult to handle. To elevate this problem, it is efficient to use DL techniques in the IoT environment. It helps for processing the data more efficiently and accurately using some DL models such as CNN, GANs, AEs. DL can process large datasets with high accuracy. Applying DL in IoT, it has the capability of performing complex data and image recognition tasks in the environment. DL reduces the time complexity and it automatically tuned new features for

210  The Smart Cyber Ecosystem for Sustainable Development Table 10.1  Datasets for deep learning in IoT. Name of the dataset

Domain

Provider

Reference

CIGAR dataset

Agriculture, climate

CCAFS

http://www.ccafs-climate.org/

Commercial Building Energy Dataset

Energy, smart building

IIITD

http://combed.github.io/

AMPds dataset

Energy, smart home

S. Makonin

http://ampds.org/

PhysioBank Databases

Healthcare

PhysioNet

https://physionet.org/ physiobank/database/

T-LESS

Industry

CMP at Czech Technical University

http://cmp.felk.cvut.cz/t-less/

Open Data Institute node Trento

Smart city

Telecom Italia

http://theodi.fbk.eu/ openbigdata/

Gas sensors for home activity monitoring

Smart home

Univ. of California San Diego

http://archive.ics.uci.edu/ml/ datasets/Gas+sensors+for+ home+activity+monitoring

MERLSense Data

Smart home, building

Mitsubishi Electric Research Labs

http://www.merl.com/wmd

T-Drive trajectory Data

Transportation

Microsoft

https://www.microsoft.com/ en-us/research/publication/ t-drive-trajectory-datasample

Traffic Sign Recognition

Transportation

K. Lim

https://figshare.com/articles/ Traffic Sign Recognition_ Testsets/4597795

RealDisp

Sport

O. Banos

http://orestibanos.com/ datasets.htm

GeoLife GPS Trajectories

Transportation

Microsoft

https://www.microsoft.com/ en-us/download/details. aspx? id=52367

Uber trip Data

Transportation

FiveThirty- Eight

https://github.com/ fivethirtyeight/ uber-tlc-foil-response

Impact of Deep Learning Techniques in IoT  211 new problems. It produces accurate results in the IoT environment. DL also contains some methods for protecting the IoT devices from malware and attacks. CNN is used for the data classification and also provides more accuracy for classifying the smart objects in the IoT environment. CNN is used to save computation time and memory space. Moreover, IoT and smart city deployment generate an enormous amount of time-series sensor data in the IoT environment. It is more efficient to use the LSTM model for this network and it is used for predicting some important events with long intervals and delays. The use of RNN in IoT devices reduces the prediction time and energy consumption in the devices and make RNN is feasible for real-time applications of IoT.

10.8 Conclusion This chapter described the DL techniques that are used in IoT. This chapter narrated about IoT architecture, characteristics, interoperability, future challenges of IoT, applications, IoT technologies, technical challenges of IoT, and its advantages. This chapter also explained DL models, applications, advantages, and disadvantages. This chapter described DL in IoT and its techniques, applications, challenges, and future directions. This chapter provided information about how DL helps to promote the IoT environment. This chapter explained the challenges and solutions that provide effective and reliable IoT systems with DL techniques. Thus, it helps the researcher to carry out their research in the upcoming technology to cope with future trends.

References 1. Abadi, M., Chu, A., Goodfellow, I., McMahan, H.B., Mironov, I., Talwar, K., Zhang, L., Deep learning with differential privacy. Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security, ACM, pp. 308– 318, 2016, October. 2. Akgul, O., Penekli, H., II, Genc, Y., Applying deep learning in augmented reality tracking. 2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS), IEEE, pp. 47–54, 2016. 3. Alom, M.Z., Taha, T.M., Yakopcic, C., Westberg, S., Sidike, P., Nasrin, M.S., Asari, V.K., A stateof-the-art survey on deep learning theory and architectures. Electronics, 8, 3, 292, 2019. 4. Alsheikh, M.A., Niyato, D., Lin, S., Tan, H.P., Han, Z., Mobile big data analytics using deep learning and apache spark. IEEE Network, 30, 3, 22–29, 2016. 5. Atzori, L., Iera, A., Morabito, G., The internet of things: A survey. Comput. Networks, 54, 15, 2787–2805, 2010. 6. Baldi, P., Autoencoders, unsupervised learning, and deep architectures, in: Proceedings of ICML workshop on unsupervised and transfer learning, pp. 37–49, 2012, June. 7. Bengio, Y., Learning deep architectures for AI, in: Foundations and trends® in Machine Learning, vol. 2, no.1, pp. 1–127, 2009. 8. Borkowski, M., Schulte, S., Hochreiner, C., Predicting cloud resource utilization. 2016 IEEE/ ACM 9th International Conference on Utility and Cloud Computing (UCC), IEEE, pp. 37–42, 2016, December. 9. Bourzac, K., Speck-size computers: Now with deep learning [news]. IEEE Spectrum, 54, 4, 13–15, 2017.

212  The Smart Cyber Ecosystem for Sustainable Development 10. Bradley, J., Barbier, J., Handler, D., Embracing the Internet of everything to capture your share of $14.4 trillion. White Paper, Cisco, 2013. 11. Chen, T., Du, Z., Sun, N., Wang, J., Wu, C., Chen, Y., Temam, O., Diannao: A small-footprint high-throughput accelerator for ubiquitous machine-learning, in: ACM Sigplan Notices, vol. 49, no. 4, pp. 269–284, ACM, 2014, February. 12. Deng, L., A tutorial survey of architectures, algorithms, and applications for deep learning. APSIPA Transactions on Signal and Information Processing, vol. 3, 2014. 13. Doersch, C., Tutorial on variational autoencoders. arXiv preprint arXiv:1606.05908, 2016. 14. Goeddel, R. and Olson, E., Learning semantic place labels from occupancy grids using CNNs. 2016 IEEE/RSJ international conference on intelligent robots and systems (IROS), IEEE, pp. 3999–4004, 2016, October. 15. Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Bengio, Y., Generative adversarial nets. Adv. Neural Inf. Process. Syst., 2672–2680, 2014. 16. Han, S., Pool, J., Tran, J., Dally, W., Learning both weights and connections for efficient neural network. Adv. Neural Inf. Process. Syst., 1135–1143, 2015. 17. He, Y., Mendis, G.J., Wei, J., Real-time detection of false data injection attacks in smart grid: A deep learning-based intelligent mechanism. IEEE Trans. Smart Grid, 8, 5, 2505–2516, 2017. 18. Hochreiter, S. and Schmidhuber, J., Long short-term memory. Neural Comput., 9, 8, 1735– 1780, 1997. 19. Datasets for machine learning, https://en:wikipedia:org/wiki/List of datasets for machine learning research, September 14, 2019. 20. Machine learning mastery, https://machinelearningmastery.com/inspirational-applicationsdeep-learning/September 18, 2019. 21. Cisco. https://www.cisco.com/c/dam/en/us/products/collateral/se/internet-of-things/at-a-glancec45-731471.pdf, September 24, 2019. 22. Edn. https://www.edn.com/5G/4442859/The-basics-of-Bluetooth-Low-Energy–BLE- September 21, 2019. 23. Verizon. https://www.verizon.com/cs/groups/public/documents/adacct/hmc_smartphone051212 _1.pdf, September 26, 2019. 24. Klaine, P.V., Imran, M.A., Onireti, O., Souza, R.D., A survey of machine learning techniques applied to self-organizing cellular networks. IEEE Commun. Surv. Tutorials, 19, 4, 2392–2431, 2017. 25. Krizhevsky, A., Sutskever, I., Hinton, G.E., Imagenet classification with deep convolutional neural networks. Adv. Neural Inf. Process. Syst., 1097–1105, 2012. 26. Lee, I. and Lee, K., The Internet of Things (IoT): Applications, investments, and challenges for enterprises. Bus. Horiz., 58, 4, 431–440, 2015. 27. Lee, J., Wang, J., Crandall, D., Šabanović, S., Fox, G., Real-time, cloud-based object detection for unmanned aerial vehicles. 2017 First IEEE International Conference on Robotic Computing (IRC), IEEE, pp. 36–43, 2017, April. 28. Li, H., Ota, K., Dong, M., Learning IoT in edge: Deep learning for the Internet of Things with edge computing. IEEE Network, 32, 1, 96–101, 2018. 29. Mohammadi, M. and Al-Fuqaha, A., Enabling cognitive smart cities using big data and machine learning: Approaches and challenges. IEEE Commun. Mag., 56, 2, 94–101, 2018. 30. Mohammadi, M., Al-Fuqaha, A., Sorour, S., Guizani, M., Deep learning for IoT big data and streaming analytics: A survey. IEEE Commun. Surv. Tutorials, 20, 4, 2923–2960, 2018. 31. Novikov, A., Podoprikhin, D., Osokin, A., Vetrov, D.P., Tensorizing neural networks. Adv. Neural Inf. Process. Syst., 442–450, 2015. 32. Ordóñez, F. and Roggen, D., Deep convolutional and lstm recurrent neural networks for multi­ modal wearable activity recognition. Sensors, 16, 1, 115, 2016.

Impact of Deep Learning Techniques in IoT  213 33. Parkhi, O.M., Vedaldi, A., Zisserman, A., Deep face recognition, in: bmvc, vol. 1, no. 3, p. 6, 2015, September. 34. Pascanu, R., Gulcehre, C., Cho, K., Bengio, Y., How to construct deep recurrent neural networks. arXiv preprint arXiv:1312.6026, 2013. 35. Patel, K.K. and Patel, S.M., Internet of things-IOT: definition, characteristics, architecture, enabling technologies, application & future challenges. Int. J. Eng. Sci. Comput., 6, 5, 2016. 36. Perera, C., Zaslavsky, A., Christen, P., Georgakopoulos, D., Context aware computing for the internet of things: A survey. IEEE Commun. Surv. Tutorials, 16, 1, 414–454, 2013.

Part 2 ARTIFICIAL INTELLIGENCE IN HEALTHCARE

11 Non-Invasive Process for Analyzing Retinal Blood Vessels Using Deep Learning Techniques Toufique A. Soomro1*†, Ahmed J. Afifi2†, Pardeep Kumar3, Muhammad Usman Keerio4, Saleem Ahmed5 and Ahmed Ali6 Electronic Engineering Department QUEST, Larkana, Pakistan Computer Vision & Remote Sensing, Technische Universität Berlin, Berlin, Germany 3 Computer Systems Engineering QUEST, Nawabshah, Pakistan 4 Electrical Engineering Department QUEST, Nawabshah, Pakistan 5 Electronic Engineering Department, Dawood University of Engineering and Technology, Karachi, Pakistan 6 Electrical Engineering Department, Sukkur IBA University, Sukkur, Pakistan 1

2

Abstract

Deep learning is an active area of research, and it demonstrated a considerable effect in diagnosing many diseases through analyzing medical images. In this book chapter, a deep learning model has been designed to analyze the retinal images for eye disease diagnosing. This research work will contain the proposal of analyzes of retinal vessels to overcome the problem of low contrast by proposing an enhancement technique as well as the segmentation of retinal blood vessels. These proposed models will help ophthalmologists analyze the network of retinal vessels by computer to recommend early treatment of eye diseases, especially diabetic retinopathy, and such a procedure has an important role in reducing the injection-based method (invasive eye diseases diagnostic method). Keywords:  Retinal color fundus images, enhancement, Principal Component Analysis (PCA), deep learning model, CNN-U-Net model, segmented retinal vessel image, sensitivity boosting, tiny vessel detections

11.1 Introduction The recent advances in medical imaging technology lead to diagnose and treat many diseases very fast. Many works have been studied the impact of automatically analyze and process medical images without expert involvement. Retina damage can be caused by many reasons such as retinal vascular occlusion and diabetic retinopathy, and the presence of pathologies in retinal fundus images and different pathologies are shown in Figures 11.1 to 11.3. In our case, accurate retinal vessel segmentation plays a significant role in diagnosing and treating patients in the early stage of the disease. The segmentation step is an important, *Corresponding author: [email protected] † Equal contribution as first author Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (217–234) © 2021 Scrivener Publishing LLC

217

218  The Smart Cyber Ecosystem for Sustainable Development

(a)

(b)

Figure 11.1  MAs in color image of the fundus image. (a) Color image of the fundus. (b) Colored background MMacula region. Remember: UTP Malaysia archive of FINDeRs is the source of these images [1, 2].

(a)

(b)

Figure 11.2  Representation of hard exudates and Soft exudates from retinal fundus images [3]. The retinal image is shown in (a) contained hard exudates. The retinal image is shown in (b) contained soft exudates. Remember: UTP Malaysia archive of FINDeRs is the source of these images [1, 2].

yet challenging task as it needs experts to perform it manually. However, different manual segmentation results may be generated from different experts and this leads to difficult disease inspection. Also, it takes a long time to segment one image and the delay of the treatment may cause blindness. In the literature [4], many computerized methods have been proposed to generate a segmented retinal blood vessel image and observe the geometric structure of the tree as shown in Figure 11.4. Different segmentation approaches have been proposed and implemented based on classical image processing algorithms. But these methods did not solve the main problems such as the detection of small vessels and, there was noise in the structure of

Analyzing Retinal Blood Vessels Using DL  219

(a)

(b)

Figure 11.3  Hemorrhages in the color fundus image. (a) Color fundus image and (b) FFFA image. Remember: UTP Malaysia archive of FINDeRs is the source of these images [1, 2].

(a)

(b)

Figure 11.4  Color images of the retinal fundus of both eyes (left and right) and it is difficult to observe the retinal vessels from these color fundus images.

retinal vessels [5] as illustrate in Figure 11.5. This chapter presents a deep learning model that generates a segmented image of blood vessels of the retina accurately. Deep Learning performs an important function in medical image observation and diagnosing disease progression. Here, we present and explain in detail a convolutional neural network (CNN) model for an accurate retinal vessel segmentation process. Our main proposal for implementing this method is to enhance the detection sensitivity of retinal blood vessels. CNN has been used for various imaging tasks, either natural images or medical images for image analysis [7]. CNN is mainly used for the classification of different feature

220  The Smart Cyber Ecosystem for Sustainable Development

Figure 11.5  Illustration of the problem of low and varying contrast in color fundus images. Note: these images are taken from the FINDeRs UTP Malaysia database [1, 6].

extraction. Various CNN models [8–10] were proposed and they are different in the depth and the building blocks according to the target task. CNNs are applied to solve various computer vision tasks such as object recognition (classification, localization, and segmentation), scene understanding (semantic segmentation and labeling), action recognition [11–14], etc. The CNN structure is based on particular loss functions to solve the particular tasks, and it is essential to select the loss function according to the properties of the images keeping an eye on the required tasks. To this end, different models have been proposed to solve the segmentation problem. They solved the segmentation task as either a dense pixelwise labeling process or from a region-based view. For retinal vessel segmentation, this task has been extensively studied and many researchers proposed different CNN models to generate fine segmented images. They apply a pre-processing step to enhance the input image. The models succeeded to recognize and segment visible large vessels, but they failed in segmenting extremely small as in Figure 11.6. The missing of small vessels has a negative impact on the performance of existing methods. We have proposed methods as well as conventional steps to detect tiny retinal vessels to improve performance over existing methods. In this research chapter, we solved each problem with a more logical and well-segmented image of the retinal vessels. First, we solved the uneven illumination and noise issue and we used the pre-processing tactics of [15, 16] and processed the pre-process database to train the CNN model. PCA and morphological tactics are used in the pre-processing, and the operation process of PCA is shown in Figure 11.7. After the training model, there are still some low contrast vessels difficult to observe, we used post-processing to get an image of fine vessels with improved performance. We have proposed a full stride-convolutional neural network with weight optimization based

Analyzing Retinal Blood Vessels Using DL  221

Figure 11.6  Problem with segmentation of retinal vessels, especially small vessels.

on the dice loss function. This work is an improved version of our previous work, and we have improved the training model to work more effectively on the dice loss function, and we have achieved better performance over existing methods. Stride-CNN is a type of encoder-decoder model with modified U-Net [17], we designed our training model as explained in the training section. Our model is based on stride convolutional layers instead of pooling layers. The retina images is an input to our model and is generated by a segmented output image with the same resolution. The key contributions of this research work are mentioned below. 1. This research work proposed a new segmentation process to obtain a fine vessels image from retinal color fundus images. 2. The Stride-CNN model is an encoder-decoder model with a modified version of U-Net with the technique of replacing all their pooling layers with a stride layer is designed for the first time for the segmentation of blood vessels of retina. 3. We used the dice loss function to obtain a segmented image of the retinal vessels. According to our knowledge, we first successfully used the dice loss function for weight optimization of the CNN model to get the retinal blood vessels.

11.2 Existing Methods Review Retinal vessel segmentation is the procedure of allocating image pixels as vessel pixels and non-vessel pixels. Many researchers [18–29] have worked on the segmentation of retinal vessels over the past two decades, but there is still room for improvement as a detailed analysis based on segmentation of retinal vessels is presented in [4, 30]. The most recent methods have given good accuracy but they have a low sensitivity which shows a lack of detection of the small vessels, as well as an improper segmentation on the pathological images [31, 32]. The proposed methods of retinal vessel segmentation are grouped into

222  The Smart Cyber Ecosystem for Sustainable Development

Irgb

IR

IG

IB

IUR

IUG

IUB

IY

ICb

ICr

Non-Uniform Illumination Removal

f (·)

PCA

v1T Iycc

v2T Iycc

v3T Iycc

λ1v1T Iycc

λ2v1T Iycc

λ3v1T Iycc

Eigenvalue Weighting

Subspace Projection

Igray

Figure 11.7  Pictorial representation of the conversion of the color image of the retinal fundus image into a gray image using Principal Component Analysis.

two classes: unsupervised and supervised techniques. The supervised techniques required the training procedure corresponding to their gold-truth, and the unsupervised methods are based on classifying the pixels into pixels of vessels or non-vessels pixels without any training process. Computerized vessel detections contain challenges such as the presence of central vessel light reflexes, bifurcations, and crossing regions between vessels, vessel fusion, low contrast, and varied contrast between vessels. Due to these problems mentioned in the segmentation of retinal vessels, as explained at the beginning, many methods have

Analyzing Retinal Blood Vessels Using DL  223 been implemented but the lack of improvement in sensitivity and is related due to the narrow-weak contrast. Few methods [28, 31, 33–36] deal with this problem and focus on the detection of small vessels. Here, the proposed deep CNN is to segment the retinal blood vessels accurately. Various methods for medical image segmentation have been proposed such as clustering, thresholding, and clustering. But deep learning has shown promising results such as the full convolution network (FCN) [13] is a dense network implemented in 2014. FCN includes convolutional layers with no fully connected layers. This network permits the segmentation process to generate the arbitrarily sized image from the input image. After FCN, many different architectures were implemented for segmentation like U-Net [17, 37]. U-Net [17] has the architecture of the encoder-decoder model for segmenting medical images. The encoder network is used to extract informative features from the input image and minimize the spatial resolution through the pooling layers. SegNet [38] is another segmentation model where the authors suggested to transfer the indices of the values in the pooling layers through the skip connection layers from the encoder network to the decoder network. This saves memory and accelerates the computations. In summary, CNN models perform well on the retinal image segmentation task, but there is still a lot of work to be done to improve the segmentation accuracy.

11.3 Methodology Automatic retinal vessel segmentation has been considered a hot research topic over the last decade. Also, applying deep learning methods in tackling computer vision and medical image analysis challenges encourages the researchers to propose and test different CNN models to solve this task. In this research work, the task is to predict and generate the vessel tree of the human eye from a retinal fundus image. This task is considered as a dense pixel classification as each pixel in the image should be classified into either a background or a vessel pixel. The final result is a generated vessel tree that is separated from the background. The main contribution of this research work is a simple, yet efficient CNN model that is a modified version of the U-Net, named Stride U-Net model. The available retinal images experience many problems like uneven illumination, low-varying contrast, and noise. It is very important to enhance these images and generate well-contrasted images before using them for training. To this end, we adopted the pre-processing technique steps in [15, 39] to enhance the training images and solve the above-mentioned problems. It is an efficient pre-processing method comprises two steps. In the first step, the uneven illumination and the noise are eliminated by applying some morphological tactics. In the second step, Principle Component Analysis (PCA) is applied to the RGB images to generate well-contrasted grayscale images. The details of the pre-processing method are explained in [15].

11.3.1 Architecture of Stride U-Net Since the breakthrough results achieved by AlexNet in the classification problem of 1K object categories, other computer vision tasks were solved using different CNN models. Mostly, the proposed CNN models consist of a chain of Convolutional Layers

224  The Smart Cyber Ecosystem for Sustainable Development (Conv-Layers) for feature extraction and pooling layers to ensure the rotation invariance and to make the network more efficient by reducing the number of parameters as well as the computation process in the network. Also, pooling layers are used to control overfitting by reducing the spatial size of the features. Also, activation functions are applied after the convolutional layers to guarantee the non-linearity in the network. Solving the retinal vessel segmentation task using deep learning is considered an image-to-image task. The model takes the pre-processed fundus image as an input image and predicts the vessel tree separated from the background as an output image. The nature of the task shows that the dense fully connected layers are not required. Figure 11.8 illustrates the details of the designed stride-CNN model. The proposed model is fed with the fundus image and predicts the vessel tree that is separated from the background. The proposed architecture has the shape of an encoder-decoder and is trained endto-end to generate the segmented image from scratch. The encoder extracts useful features and forwards them to the decoder that reconstructs the image once more and produces the segmentation map image. Pixels in the output image are classified as either foreground pixels (vessels) or background pixels (non-vessels). The encoder network comprises five convolutional blocks. The convolutional block composed of three convolutional layers and rectified linear units (ReLU) to apply a nonlinearity operation on the features. For the down sampling of the features’ spatial resolution, strided convolutional layers have been used to replace the max-pooling layers that are normally used for this purpose. The strided convolutional layers are trainable layers that extract useful information for better segmentation map images. The convolutional layers have a kernel with a size of 3 × 3 and a stride of 1. The strided convolutional layers have a kernel size of 2 × 2 and a stride of 2. The feature maps generated from each block have a size of 16, 32, 64, 128, and 256, respectively.

conv(3,3,1) + conv(3,3,1) + conv(3,3,1) + strided_conv(2,2,2) d_conv(3,3,1) + conv(3,3,1) + conv(3,3,1)

concat & fuse

Figure 11.8  The proposed CNN architecture. The encoder network and the decoder network consist of five convolutional blocks. conv(a1, a2, a3) denotes the convolutional layer that has a kernel size of a1 × a2 and with a stride of a3. d-conv(b1, b2, b3) denotes the deconvolutional layer with a kernel size of b1 × b2 and a stride of b3.

Analyzing Retinal Blood Vessels Using DL  225 The decoder network is responsible for taking the useful features from the encoder network and reconstruct the image to produce the segmented vessel images. The decoder architecture is almost similar to the encoder, but with more layers for the purpose of features concatenation and fusion. It has five convolutional blocks. Each block has a deconvolutional layer, to upsample the features and a ReLU as an activation layer. The deconvolutional layers increase the resolution of the features. Also, it has two convolutional layers with a kernel size of 3 × 3 and a stride of 1. To infer the finer segmentation map image with distinguished edges of the vessels, the features from the encoder are transferred and combined with their correspondences in the decoder. This is performed using the skip connections between the two networks. The features extracted in the encoder network have rich information and they preserve the spatial details of the edges. So, transferring them to the decoder and fuse them with the corresponding features enhance the last output and segment more vessels accurately.

11.3.2 Loss Function After preparing the training data and design the CNN model, we have to select the proper loss function for the training of the model and for the optimization of the model parameters. This step is very important as it affects the performance of the model. The retinal vessel segmentation task is defined to categorize every pixel in the input image being non-vessel or a vessel. Statistically, the vessel pixels in the input image are almost 10% of the whole pixels, and the remaining 90% of the pixels are considered as a background. That is, the distribution of the classes is highly imbalanced. For better training process and not to be trapped in local minima, this issue should be considered. Otherwise, the vessels will be misclassified or partially identified. To this end, a suitable loss function should be selected to help in solving this critical issue. Here, we decide to employ the Dice Loss function for the optimization of the model parameters. The network is designed to generate a two-channel image. Each channel represents either a foreground class or a background class. The two values of each pixel denote the probability of the pixel being belonging to one of the classes. A differentiable approximation of the loss function is specified as follows:

Ldice = 1 −

∑

2

∑

x ∈Ω

2 l

x ∈Ω

pl ( x ) g l ( x )

p (x ) +

∑

x ∈Ω

2 l

g (x )

(11.1)



In the equation, pl (x) is the score of pixel x belonging to class l and gl (x) is the groundtruth label vector where it is 1 on the true class and 0 for the other class. As mentioned before, the problem suffers from imbalanced data and this loss function helps in overcoming this issue during training. There are no weighting parameters in the function of the classes.

11.4 Databases and Evaluation Metrics Publicly accessible retinal image databases are used to validate our method and compare it with the state of the art. The first database is called DRIVE (Digital Retinal Images for

226  The Smart Cyber Ecosystem for Sustainable Development Table 11.1  Evaluating parameters. Parameters

Definition

Sensitivity (SEN)

Sensitivity is a statistical parameter that calculates the percentage of true positive values (these are vessels pixels) that are segmented.

Specificity (SPE)

Specificity is a statistical parameter that calculates the percentage of true negative values (these are non-vessels pixels) that are segmented.

Accuracy (ACC)

Area Under Curve (AUC)

Accuracy is a parameter that gives overall correctly classified pixels information about the vessels, and it is the sum of the total pixels of the vessels and the total number of the non-vessel pixels up to the aggregates of the total number of pixels. Vessels imbalance classifications mislead performance, and the area under the curve gives information on non-vessels and vessels pixels. AUC is measured by comparing the performance of the classification of true vessels (this is the sensitivity) and the true classification of nonvessels (this is the specificity).

Mathematical representation Sensitivity

TP TP FN

Specificity

TN TN FP

Accuracy

TP

TP TN FP FN TN

AUC = (Se + Sp)/2

Vessel Extraction) [19] and it contains 40 images with their corresponding ground-truth. The DRIVE database has an image resolution of 768 × 584 pixels. The second database is called STARE (Structured Retinal Analysis Database) [40] which contains 20 retinal images with ground-truth. The images at 50% of the STARE databases contain images on pathologies and allow us to test the capacity of the segmentation method. All the researchers used these databases and it gives us the opportunity to relate our techniques’ results with traditional techniques. The performance of the method is measured through four parameters: accuracy, area under curve, specificity, and sensitivity. The description of these parameters is given in Table 11.1.

11.4.1 CNN Implementation Details We implement the proposed model and train it based on MatConvNet [41]. Initialization of model parameters such as model weight is initialized using Xavier’s initialization process and we trained our CNN model from scratch. The basis of the Stochastic Gradient Descent process on the backpropagation network is used to optimize the metrics of the network

Analyzing Retinal Blood Vessels Using DL  227 and solve the optimization problem and mathematically it is expressed as Equation 11.1. The training hyper-parameters are initialized as follows: the learning rate is set to 10−3 and it decreases during training. The momentum and the weight decay are set and frozen as 0.9 and 10−5, respectively. We run the training process and stop it once there is no change in the model weights. To increase the training data, we applied the data augmentation technique to the original training images. We extract small patches from the pre-processed images and resized them to the original image spatial resolution. The extracted patches are overlapped and they are used besides the complete images. This makes the network to generalize more and learns different vessels size.

11.5 Results and Analysis 11.5.1 Evaluation on DRIVE and STARE Databases We have experimented our model on the DRIVE databases and the STARE database and achieved good performance with a sensitivity of 82.24% on the DRIVE database and 81.3% on the STARE database as shown in Table 11.2. The results obtained recommend that the CNN model be marketed in eye hospitals for appropriate analysis of retinal blood vessels. Along with the quantitative analysis, we perform a visualization analysis of our CNN model because the segmented vessel images are shown in Figure 11.9 with their ground-truth images.

11.5.2 Comparative Analysis Comparative is one of the main steps for evaluation of the method capability. We are performing a comparative analysis of our CNN model with deep learning techniques and other methods built entirely based on the traditional image processing tactics on the DRIVE and STARE databases. Table 11.3 represents the comparative analysis of our CNN model with deep learning methods, and our model outperforms other deep learning methods with the highest sensitivity and comparable accuracy. Table 11.4 presented the comparative analysis of our method with other traditional image processing methods for the segmentation of retinal blood vessels, and our implemented model outperforms other methods in overall performance. The comparative analysis of the model shows that the designed model is able to segment the image of retinal vessels and that it will be an excellent tool for ophthalmologists to diagnose eye diseases and recommend to the patient for early treatment.

Table 11.2  Evaluation on databases. Database

Sen (%)

Spe (%)

ACC (%)

AUC (%)

DRIVE STARE

82.4 81.3

96.2 95.9

95.1 94.9

85.8 85.2

228  The Smart Cyber Ecosystem for Sustainable Development

Figure 11.9  Qualitative results of the retinal vessel segmentation. From top to bottom: input image, groundtruth, and the segmented image by the proposed model.

Table 11.3  Comparative analysis: CNN methods. Note: All parameters are measured as percentage. Database

DRIVE

STARE

Methods

SEN

SPE

ACC

AUC

SEN

SPE

ACC

AUC

Zhang et al. [42] Maji et al. [43] Liskowski et al. [44] Fu et al. [45] Wu et al. [46] Yao et al. [47] Maninis et al. [48] Fu et al. [49] Tan et al. [50] M et al. [51] Song et al. [52] Soomro et al. [15] Brancati et al. [53] Yan et al. [54] Soomro et al. [55] Proposed Method

76.0 77.3 0.729 75.3 66 75 74.6 74.2 76.5 73.9 82.4

96 96.9 98.5 97.9 91.7 98.2 98.1 95.6 96.2

94.0 94.7 94.9 95.2 97 93.6 94.7 92.6 95.6 94.9 94.8 95.4 95.4 94.8 95.1

97.3 -

74.1 71.4 74.8 75.8 74.8 81.3

92.2 98.4 96.2 95.9

94.9 95.8 95.4 94.7 96.1 94.7 94.9

98.2 83.1 83.5 98.1 85.5 85.2

82.2 83.1 97.5 84.4 85.8

Analyzing Retinal Blood Vessels Using DL  229 Table 11.4  Comparative analysis on existing methods. Note: All parameters are measured as a percentage. Database

DRIVE

STARE

Methods

SEN

SPE

ACC

AUC

SEN

SPE

ACC

AUC

Staal et al. [19] Soares et al. [20] Mendonca et al. [24] Martinez-Perez et al. [56] Al-Diri et al. [25] Lupas et al. [57] Palomera-Perez et al. [58] Xinge et al. [21] Marin et al. [22] Fraz et al. [26] Nguyen et al. [27] Orlando et al. [59] Yin et al. [60] Roychowdhury et al. [61] Melinscak et al. [62] Annunziata et al. [63] Li et al. [64] Zhao et al. [65] Soomro et al. [66] Khan et al. [33] Zhang et al. [67] Orlando et al. [68] Ngo et al. [69] Thangaraj et al. [70] Biswal et al. [71] Soomro et al. [31] Soomro et al. [72] Proposed Method

73.4 72.4 72.8 72 66 74.1 70.6 74.1 78.5 72.5 75.6 71.6 71.3 73.4 74.3 78.9 74.6 80.1 95 75.2 74.5 82.4

97.6 96.5 95.5 96.1 97.5 98 98.1 96.7 98.3 98.1 97.8 96.8 96.7 97.6 96.8 98.4 97.5 97 97.6 96.2 96.2

94.6 94.6 94.5 93.4 95.9 92.2 94.3 94.5 94.8 94 94.7 95.2 94.6 95.2 94.4 94.1 95.1 94.7 95.3 96.1 71 95.3 94.8 95.1

85.5 84.5 84.2 81.1 85.8 84.3 97.4 96.2 97.4 97.4 84.8 84.1 85 95.2 97.5 88.8 85.8

69.9 75 75.2 77.9 72.6 69.4 75.4 77.2 71.3 77.3 77.6 71.1 73.6 76.7 76.8 83.4 70 78.6 78.4 81.3

97.3 95.6 96.8 94 97.5 98.1 97.3 97.3 98.4 98.4 95.4 96.5 97.1 97.6 97.3 95.3 97 98.2 97.6 95.9

95.1 94.8 94.4 94.1 92.4 94.9 95.2 95.3 93.2 95.1 95.1 95.6 96.2 94.3 94.2 95 95.4 94.4 95 96.7 95.1 94.9

83.6 85.3 86 86 85.1 83.8 97.7 96.9 96.5 98.7 86.5 83.8 85.3 96.1 89.4 85.2

11.6 Concluding Remarks The retinal blood vessel segmentation from digital retinal images is an active research topic in recent years. Several approaches based on traditional image processing tactics or machine learning are implemented to enhance the segmentation accuracy of the blood vessels of the retina. Because accurate detection of the blood vessels of the retina helps ophthalmologists observe the progress of the disease. In this research book chapter, we have presented the Stride CNN model that is inspired by U-Net. The modification is based on replacing all grouping layers with Stride convolutional layers. The proposed model takes the pre-processed image and produces the segmented vessel images with the same input image spatial resolution. We used the dice loss feature to get accurate retinal vessels. Our method outperforms other techniques. Still, there are many challenges to tackle, the authors only used two databases, and the method can be validated further on different databases from

230  The Smart Cyber Ecosystem for Sustainable Development different regions of the world. We can further improve the performance by testing other loss functions that help to get train data to get better performance.

References 1. Hani, A. and Soomro, T.A., Non-invasive contrast enhancement for retinal fundus imaging. IEEE International Conference on Control System, Computing and Engineering (ICCSCE), vol. 1, pp. 197–202, 2013. 2. Soomro, T.A., Non-invasive image denoising and contrast enhancement techniques for retinal fundus images, M.S. thesis, Department of Electrical and Electronic Engineering, University Teknologi PETRONAS, Malaysia, 2014. 3. Faust, O. and Acharya et al., R., Algorithms for the automated detection of diabetic retinopathy using digital fundus images: A review. J. Med. Syst., 36, 1, 145–157, 2010. 4. Soomro, T.A., Gao, J., Khan, T.M., Hani, A.F.M., Khan, M.A.U., Paul, M., Computerised approaches for the detection of diabetic retinopathy using retinal fundus images: A survey. J. Pattern Anal. Appl., 20, 1–35, 2017. 5. Soomro, T.A., Gao, J., Khan, M.A.U., Khan, T.M., Paul, M., Role of image contrast enhancement technique for ophthalmologist as diagnostic tool for diabetic retinopathy. 2016 International Conference on Digital Image Computing: Techniques and Applications (DICTA), Gold Coast, QLD, pp. 1–8, 2016. 6. Toufiq, Non-Invasive Image Processing and Deep Learning Methods for Retinal Vessels Segmentation for Early Detection of Eye Diseases, Ph.D. thesis, School of Computing and Mathematics, CSU, NSW, Australia, 2018. 7. Krizhevsky, A., Sutskever, I., Hinton, G.E., Imagenet classification with deep convolutional neural networks. In Conference on Advances in Neural Information Processing Systems, pp. 1097–1105, 2012. 8. Simonyan, K. and Zisserman, A., Very deep convolutional networks for large-scale image recognition. ICLR, 2015. 9. Szegedy, C., Liu, W., Jia, Y., Sermanet, Y., Reed, P., Anguelov, S., Erhan, D., Vanhoucke, D., Rabinovich, V., Going deeper with convolutions. CVPR, 2015. 10. He, K., Zhang, X., Ren, S., Sun, J., Deep residual learning for image recognition. Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778, 2016. 11. Afifi, A.J., Hellwich, O., Soomro, T.A., Simultaneous object classification and viewpoint estimation using deep multi-task convolutional neural network. VISIGRAPP, 2018. 12. Afifi, A.J. and Hellwich, O., Object depth estimation from a single image using fully convolutional neural network. 2016 International Conference on Digital Image Computing: Techniques and Applications (DICTA), pp. 1–7, Nov 2016. 13. Shelhamer, E., Long, J., Darrell, T., Fully convolutional networks for semantic segmentation. IEEE transactions on pattern analysis and machine intelligence, vol. 39, no. 4, pp. 640–651, 2017. 14. Khan, M.A.U., Khan, T.M., Aziz, K., II, Ahmad, S.S., Mir, N., Elbakush, E., The use of Fourier phase symmetry for thin vessel detection in retinal fundus images. 2019 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT), IEEE, pp. 1–6, 2019. 15. Soomro, T.A., Afifi, A.J., Gao, J., Hellwich, O., Khan, M.A.U., Paul, M., Zheng, L., Boosting sensitivity of a retinal vessel segmentation algorithm with convolutional neural network. International Conference on Digital Image Computing: Techniques and Applications (DICTA), 2017. 16. Khan, M.A.U., Khan, T.M., Bailey, D.G., Kittaneh, O., Deriving scale normalisation factors for a glog detector. IET Image Proc., 12, 9, 1673–1682, 2018.

Analyzing Retinal Blood Vessels Using DL  231 17. Ronneberger, O., Fischer, P., Brox, T., U-net: Convolutional networks for biomedical image segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention Springer, Cham, vol. 1, pp. 234–241, 2015. 18. Zhang, M., Li, X., Yang, Z., Yang, Y., A novel zero-crossing edge detection method based on multi-scale space theory. IEEE 10th International Conference on Signal Processing, vol. 1, pp. 1036–1039, 2010. 19. Staal, J., Abramoff, M.D., Niemeijer, M., Viergever, M.A., van Ginneken, B., Ridge based vessel segmentation in color images of the retina. IEEE Trans. Med. Imaging, 23, 4, 501–509, April 2004. 20. Soares, J.V.B., Leandro Roberto, J.J.G., Cesar, M., Jelinek, H.F. Jr, Cree, M.J., Retinal vessel segmentation using the 2-d gabor wavelet and supervised classification. IEEE Trans. Med. Imaging, 25, 9, 1214–1222, 2006. 21. Xinge, Y., Qinmu, P., Yuan, Y., Yiu-ming, C., Jiajia, L., Segmentation of retinal blood vessels using the radial projection and semi- supervised approach. Pattern Recognit., 44, 10–11, 2011. 22. Marin, D., Aquino, A., Gegundez-Arias, M.E., Bravo, J.M., A new supervised method for blood vessel segmentation in retinal images by using gray-level and moment invariants-based features. IEEE Trans. Med. Imaging, 30, 1, 146–158, 2011. 23. Wang, Y., Ji, G., Lin, P., Trucco, E., Retinal vessel segmentation using multiwavelet kernels and multiscale hierarchical decomposition. Pattern Recognit., 46, 8, 2117–2133, 2013. 24. Mendonca, A.M. and Campilho, A.A., Segmentation of retinal blood vessels by combining the detection of centerlines and morphological reconstruction. IEEE Trans. Med. Imaging, 25, 1200–1213, 2006. 25. Al-Diri, B., Hunter, A., Steel, D., An active contour model for segmenting and measuring retinal vessels. IEEE Trans. Med. Imaging, 28, 9, 1488–1497, 2009. 26. Fraz, M.M., Remagnino, P., Hoppe, A., Uyyanonvara, B., Rudnicka, A.R., Owen, C.G., Barman, S.A., An ensemble classification-based approach applied to retinal blood vessel segmentation. IEEE Trans. Biomed. Eng., 59, 9, 2538–2548, 2012. 27. Nguyen, U.T.V., Bhuiyan, A., Park, L.A.F., Ramamohanarao, K., An effective retinal blood vessel segmentation method using multi-scale line detection. Pattern Recognit., 46, 703–715, 2013. 28. Soomro, T.A., Khan, M.A.U., Junbin, Khan, T.M., Paul, M., Contrast normalization steps for increased sensitivity of a retinal image segmentation method. J. Signal Image Video Process., 11, 1509–1517, 2017. 29. Soomro, T.A., Afifi, A.J., Ali Shah, A., Soomro, S., Baloch, G.A., Zheng, L., Yin, M., Gao, J., Impact of image enhancement technique on cnn model for retinal blood vessels segmentation. IEEE Access, 7, 158183–158197, 2019. 30. Soomro, T.A., Afifi, A.J., Zheng, L., Soomro, S., Gao, J., Hellwich, O., Paul, M., Deep learning models for retinal blood vessels segmentation: A review. IEEE Access, 7, 71696–71717, 2019. 31. Soomro, T.A., Khan, T.M., Khan, M.A.U., Gao, J., Paul, M., Zheng, L., Impact of ica-based image enhancement technique on retinal blood vessels segmentation. IEEE Access, 6, 3524–3538, 2018. 32. Soomro, T.A., Afifi, A.J., Gao, J., Hellwich, O., Zheng, L., Paul, M., Strided fully convolutional neural network for boosting the sensitivity of retinal blood vessels segmentation. Expert Syst. Appl., 134, 36 – 52, 2019. 33. Khan, M.A.U., Soomro, T.A., Khan, T.M., Bailey, D.G., Gao, J., Mir, N., Automatic retinal vessel extraction algorithm based on contrast-sensitive schemes. International Conference on Image and Vision Computing New Zealand (IVCNZ), pp. 1–5, Nov 2016. 34. Bailey, D.G., Khan, M.A.U., Khan, T.M., Soomro, T.A., A generalized multi-scale line-detection method to boost retinal vessel segmentation sensitivity. Pattern Anal. Appl., 22, 1–20, 2018. 35. Soomro, T.A., Mir, N., Khan, M.A.U., Khan, T.M., Gao, J., Boosting sensitivity of a retinal vessel segmentation algorithm. Pattern Anal. Appl., 22, 1–17, 2017.

232  The Smart Cyber Ecosystem for Sustainable Development 36. Khan, M.A.U., Khan, T.M., Mir, N., Soomro, T.A., Gao, J., Boosting sensitivity of a retinal vessel segmentation algorithm. Pattern Anal. Appl., 22, 2, 583–599, 2019. 37. Khan, M.A.U., Khan, T.M., Naqvi, S.S., Khan, M.A., Ggm classifier with multi-scale line detectors for retinal vessel segmentation. Signal Image Video Process., 13, 8, 1667–1675, 2019. 38. Badrinarayanan, V., Handa, A., Cipolla, R., Segnet: A deep convolutional encoder-decoder architecture for robust semantic pixel-wise labelling. arXiv preprint arXiv:1505.07293, 1–10, 2015. 39. Soomro, T.A., Paul, M., Gao, J., Zheng, L., Retinal blood vessel extraction method based on basic filtering schemes. 2017 IEEE International Conference on Image Processing (ICIP), pp. 4422–4426, 2017. 40. Hoover, A.D., Kouznetsova, V., Goldbaum, M., Locating blood vessels in retinal images by piecewise threshold probing of a matched filter response. IEEE Trans. Med. Imaging, 19, 3, 203–210, 2000. 41. Vedaldi, A. and Lenc, K., Matconvnet: Convolutional neural networks for matlab. 23rd ACM International Conference on Multimedia, pp. 689–692, 2015. 42. Zhang, J., CuiWuhan, Y., Wang, J.L., Blood vessel segmentation of retinal images based on neural network. Image and Graphics. ICIG 2015. Lecture Notes in Computer Science, vol. 9218, pp. 11–17, 2015. 43. Maji, D., Santara, A., Mitra, P., Sheet, D., Ensemble of deep convolutional neural networks for learning to detect retinal vessels in fundus images. arXiv:1603.04833v1 [cs.LG], 1–4, 2016. 44. Liskowski, P. and Krawiec, K., Segmenting retinal blood vessels with deep neural networks. IEEE Trans. Med. Imaging, 35, 11, 2369–2380, 2016. 45. Fu, H., Xu, Y., Wong, D.W.K., Liu, J., Retinal vessel segmentation via deep learning network and fully-connected conditional random fields. International Symposium on Biomedical Imaging (ISIB), pp. 10–13, 2016. 46. Wu, A., Xu, Z., Gao, M., Buty, M., Mollura, D.J., Deep vessel tracking: A generalized probabilistic approach via deep learning. IEEE 13th International Symposium on Biomedical Imaging (ISBI), vol. 1, pp. 1363–1367, 2016. 47. Yao, Z., Zhang, Z., Xu, L.Q., Convolutional neural network for retinal blood vessel segmentation. 9th International Symposium on Computational Intelligence and Design (ISCID), vol. 1, pp. 406–409, 2016. 48. Maninis, K.-K., Pont-Tuset, J., Arbelaez, P., Van Gool, L., Deep retinal image understanding. In Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 140–148, 2016. 49. Fu, H., Xu, Y., Lin, S., Wing, D., Wong, K., Liu, J., Deepvessel: Retinal vessel segmentation via deep learning and conditional random field. International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 132–139, 2016. 50. Tan, J.H., Rajendra Acharya, U., Bhandary, S.V., Chua, K.C., Sivaprasad, S., Segmentation of optic disc fovea and retinal vasculature using a single convolutional neural network. J. Comput. Sci., 20, 70–79, 2017. 51. Frucci, M., Riccio, D., Sanniti di Baja, G., Serino, L., Direction-based segmentation of retinal blood vessels. Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications. CIARP. Lecture Notes in Computer Science Springer, Cham, vol. 10125, pp. 1–8, 2017. 52. Song, J. and Lee, B., Development of automatic retinal vessel segmentation method in fundus images via convolutional neural networks. 39th IEEE Annual International Conference of the Engineering in Medicine and Biology Society (EMBC), pp. 681–684, 2017. 53. Brancati, N., Frucci, M., Gragnaniello, D., Riccio, D., Retinal vessels segmentation based on a convolutional neural network. Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications. CIARP. Lecture Notes in Computer Science Springer, Cham, vol. 10657, pp. 119–126, 2018.

Analyzing Retinal Blood Vessels Using DL  233 54. Yan, Z., Yang, X., Tim Cheng, K.-T., Joint segment-level and pixel-wise losses for deep learning based retinal vessel segmentation. IEEE Trans. Biomed. Eng., 65, 1–10, 2018. 55. Soomro, T.A., Afifi, A.J., Gao, J., Hellwich, O., Manoranjan Paul, and Lihong Zheng, “Strided u-net model: Retinal vessels segmentation using dice loss. Digital Image Computing: Techniques and Applications (DICTA), pp. 1–8, 2018. 56. Martinez-Perez, Hughes, A.D., Thom, S.A., Segmentation of blood vessels from red-free and fluorescein retinal images. J. Med. Image Anal., 11, 1, 47–61, 2007. 57. Lupas, C.A., Tegolo, D., Trucco, E., Fabc:retinal vessel segmentation using adaboost. IEEE Trans. Inf. Technol. Biomed., 14, 5, 1267–1274, 2010. 58. Palomera-Perez, M.A., Martinez-Perez, M.E., Benitez-Perez, H., Ortega-Arjona, J.L., Parallel multiscale feature extraction and region growing: Application in retinal blood vessel detection. IEEE Trans. Inf. Technol. Biomed., 14, 2, 500–506, 2010. 59. Orlando, J., II and Blaschko, M., Learning fully-connected crfs for blood vessel segmentation in retinal images. Conference on Medical Image Computing and Computer-Assisted Intervention, vol. 17, pp. 634–641, 2014. 60. Yin, X., Ng, B.W.-H., He, J., Zhang, Y., Abbott, D., Accurate image analysis of the retina using hessian matrix and binarisation of thresholded entropy with application of texture mapping. PLOS ONE, 9, 4, 1–17, 2014. 61. Roychowdhury, S., Koozekanani, D.D., Parhi, K.K., Blood vessel segmentation of fundus images by major vessel extraction and subimage classification. IEEE J. Biomed. Health Inf., 19, 03, 1118–1128, 2015. 62. Melinscak, M., Prentasic, P., Loncaric, S., Retinal vessel segmentation using deep neural networks. International Confernence on Computer Vision Theory and Application, pp. 1–6, 2015. 63. Annunziata, R., Garzelli, A., Ballerini, L., Mecocci, A., Trucco, E., Leveraging multiscale hessian-based enhancement with a novel exudate inpainting technique for retinal vessel segmentation. IEEE J. Biomed. Health Inf., 20, 4, 1129–1138, 2016. 64. Li, Q., Feng, B., Xie, L.P., Liang, P., Zhang, H., Wang, T., A cross-modality learning approach for vessel segmentation in retinal images. IEEE Trans. Med. Imaging, 35, 01, 109–118, 2016. 65. Zhao, Y., Rada, L., Chen, K., Harding, S.P., Zheng, Y., Automated vessel segmentation using infinite perimeter active contour model with hybrid region information with application to retinal images. IEEE Trans. Med. Imaging, 34, 9, 1797–1807, 2015. 66. Soomro, T.A., Khan, M.A.U., Gao, J., Khan, T.M., Paul, M., Mir, N., Automatic retinal vessel extraction algorithm. International Conference on Digital Image Computing: Techniques and Applications (DICTA), pp. 1–8, Nov 2016. 67. Zhang, J., Dashtbozorg, B., Bekkers, E., Pluim, J.P.W., Duits, R., ter Haar Romeny, B.M., Robust retinal vessel segmentation via locally adaptive derivative frames in orientation scores,”. IEEE Trans. Med. Imaging, 35, 12, 2631–2642, 2016. 68. Orlando, J.I., Prokofyeva, E., Blaschko, M.B., A discriminatively trained fully connected conditional random field model for blood vessel segmentation in fundus images. IEEE Trans. Biomed. Eng., 64, 1, 16–27, 2017. 69. Ngo, L. and Han, J.H., Multi-level deep neural network for efficient segmentation of blood vessels in fundus images. Electron. Lett., 53, 16, 1096–1098, 2017. 70. Thangaraj, S., Periyasamy, V., Balaji, R., Retinal vessel segmentation using neural network. IET Image Proc., 12, 5, 669–678, 2018. 71. Biswal, B., Pooja, T., Bala Subrahmanyam, N., Robust retinal blood vessel segmentation using line detectors with multiple masks. IET Image Proc., 12, 3, 389–399, 2018. 72. Soomro, T.A., Gao, J., Lihong, Z., Afifi, A.J., Soomro, S., Paul, M., Retinal blood vessels extraction of challenging images. Data Mining. AusDM 2018. Communications in Computer and Information Science, Springer, Singapore, vol. 996, no. 1–12, 2019.

12 Existing Trends in Mental Health Based on IoT Applications: A Systematic Review Muhammad Ali Nizamani, Muhammad Ali Memon and Pirah Brohi* Faculty of Engineering & Technology, University of Sindh Jamshoro, Sindh, Pakistan

Abstract

Today, Internet of Things (IoT) is one of the fastest growing field along with the Artificial Intelligence. It connects different wearable embedded devices (things) having unique identities with the internet and helps in ubiquitous data collection and analysis. Its applications range in wide variety of areas from smart ambient houses to smart agriculture. One of the important emerging area of application is the role IoT can play in improving personal healthcare, especially the mental healthcare of persons, where limited number of mental health practitioners means limited care for mental illness related treatment. In the last decade, IoT revolution has rebuild the modern healthcare with the promise of good healthcare services with timely intervention. In this work, a review of the existing trends in mental healthcare based on IoT applications is performed. In addition, this work reviews the benefits of IoT applications in healthcare and challenges it needs to address for widespread use of IoT in mental healthcare. For example, one of the challenges IoT-based mental healthcare applications face is that of security and trust, for which a blockchain-based approach in IoT is highlighted in this paper. Keywords:  Human computer interaction, mental health, IoT healthcare applications, blockchain in IoT

12.1 Introduction Internet of Things (IoT) is based on correlated computing technologies, with unique identities that can transmit the data over a network without demanding any human involvement. The purpose of IoT was to decrease the human exertions with smarter devices. These smarter devices move each and every data of daily life with the help of sensors to process and complete a task. Things in IoT are the physical objects in real world, such as wearable sensors and embedded devices, which are joined through communication network. It works like a bridge between real world and the digital world [1, 2]; Figure 12.1 shows the general architecture for IoT. In general, many sensor nodes, the things, are connected with IoT gateway. IoT gateway can directly connect to Internet/WAN or connect using edge router, which then connects with a cloud, or a back server or with remote devices, such as mobile phones using Apps. IoT has huge potential and is being used in vastly different areas, *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (235–250) © 2021 Scrivener Publishing LLC

235

236  The Smart Cyber Ecosystem for Sustainable Development for example, in agriculture, smart warehouses, defense, and ambient intelligence in smart homes. One of the areas where IoT has great ability to improve performance and services is the delivery of healthcare. Healthcare-based IoT is the one of the fastest emerging application of the IoT [3]. IoT has the capacity to deliver and provide advancement in several medical applications such as remote health monitoring system, with cure and prescription at home and by healthcare providers. Thus, the innumerable medical expedients, sensors, and diagnostic and imaging devices are viewed as smart technologies being fundamental to the IoT in personal well-being and it will lead to minimize the cost of healthcare and at the same time provide efficient patient health monitoring [4, 5]. The number of healthcare workers are already low in low-income countries, and furthermore, the mental health practitioners are very low, as low as below 1 per 100,000 population, compared to the global median of 9 (even in high-income countries, the figure is 72, not that high) [6]. Societies are increasing the attention on mental health well-being along with physical well-being using modern tools [7]. IoT technologies have enhanced the traditional diagnostic apparatuses of the last decade, like magnetic resonance imaging, epigenetic, bionics, and neuropsychological tests [8]. Bionics is enhanced by IoT sensors to get patient’s data to proficiently identify the psychiatric diseases with the help of big data analysis [9]. IoT suggests very smart pioneering solutions to encourage the supplementary methodologies in the area of AHA (Active and Healthy Ageing for the elderly persons). It has enlarged as a key of empowering the wide range of technologies to supports the users to take care of their health by themselves. Patient’s health monitoring status applications are the examples of how IoT is involved in improving patient quality of life and supporting in healthy ageing. Figure 12.2 shows the basic system architecture for the IoT in the healthcare. Globally, there is large population of person is suffering from some form of mental illness, such as depression, dementia (Alzheimer’s disease), and bipolar disorder, and many children are suffering from the autism [10]. These mental illnesses have a huge influence on patient’s life as well as his/her economic status [11]. In that manner, a wide range of variety

Internet Cloud Server

Temperature Pressure Humidity Mic Camera Sensor Nodes

Figure 12.1  IoT architecture.

IoT Gateway/Framework

Edge Router

Remote Device/Mobile

Trends in Mental Health Based on IoT  237 Business Application Notification

Reports

Check List

Storage

Data Integration

Cloud Storage

Machine to Machine Integration

IoT Connectivity Layer

Router Gateway

Medical Sensor Nodes Blood pressure Body temperature monitor monitor

Cardiac monitor

Figure 12.2  Basic architecture for IoT in healthcare.

of IoT-based applications are enabled for the solution of better healthcare in these types of mental illness. In the healthcare, background information from networked sensors have an optimistic transformation of improvement [12]. In this work, the state of art in usage of smart technology for enhancement of patient mental health, where IoT applications play very important role is analyzed and also given is an overview of related IoT services. There are analogous reviews that focus their research on the challenges of IoT in Healthcare system [13], along with exploration of barriers of patient’s mental health monitoring solutions adopted by clinicians in healthcare organization [14]. Following are the sections that will be discussed in this paper; the methodology that has been used to survey and select research work for IoT in the mental health, then applications in IoT for mental health and the benefits of IoT in mental health are given, Also, discussed are the challenges faced for the large-scale deployment of IoT in mental healthcare and the role of blockchain in building trust in IoT for healthcare in general. Subsequently, the results of IoT application on mental health well-being are analyzed, and then, survey work is ended with the discussion, description of limitation of the work, and a general conclusion.

12.2 Methodology This paper presents a survey based on existing trends in mental health applications based on IoT. The survey search is based on papers published in between 2016 and 2020 in English. This paper discusses research studies, those are based on mental health applications based on IoT and challenges related to mental health applications. Furthermore, this paper contains the articles, those are demonstrating the benefits of IoT-based mental healthcare

238  The Smart Cyber Ecosystem for Sustainable Development applications. We have conducted survey papers and articles to classify these papers. We have explored four online databases such as IEEE Xplore, Elsevier (ScienceDirect), Google Scholar, and PubMed. Searching keywords were mental healthcare applications, challenges and benefits of IoT-based applications, or IoT in mental healthcare. For the benefits and challenges for IoT-based mental healthcare applications and services, a number of studies were surveyed. After that, related main articles were deliberated. IoT is expanding highly and with more and more devices being connected through IoT and related computing being performed is becoming complex and needs to be context aware, whether done: using a simple smart watch, edge computing, fog computing, or done on the cloud.

12.3 IoT in Mental Health IoT has a wide-ranging applicability in mental health. Mental health is denoting as cognitive, behavioral, and emotional well-being. The mental health has an influence on physical health as well and it can also influence on people’s daily life and on their relationships. Mental health is consisting of a one’s ability of enjoying his/her life to accomplish the stability of life activities and struggles to accomplish the psychological resilience. With the busy lifestyle, mental health is habitually neglected by most of us. But this, in turn, upsets their physical health as well [15]. For physical illness here are so many signs that show physical illness, while the signs of mental illness are regularly ignored. Ultimately, this increases danger of mental exhaustion which comes to be a cause for stress, depression, sleep disorders, etc. [16]. The major cure for mental diseases are psychotherapy, medication such as anti-depressants and anti-anxiety medication and mood stabilizers, etc., in that manner, by using IoT-based technologies, mental healthcare provider organizations can expand the excellence of a good mentally healthy life. Furthermore, IoT-based monitoring system can provide the advantage to doctors in the cures and expect a symptom before beginning the identification of the diseases [17]. A basic architecture for IoT for mental health includes wearable devices for monitoring different mental health aspects, such as monitoring for dementia related issues. For example, a mentally ill patient losses way and forgets about his home and moves to a path that is away from his home location, the wearable device using context can alert the family, care givers, and the police. All this requires careful collection of data by sensors using context computing, which can be done remotely on cloud or on local computer and then if an alarm is generated can be transmitted to family or caregivers mobile phone, etc., via internet; Figure 12.3 shows this scenario.

12.4 Mental Healthcare Applications and Services Based on IoT The field of IoT has been growing as innovative area. Billions of devices are connected with each other to Internet over several access networks supported by tools [18]. The growing use of IoT for patient health related monitoring helps quickly observe the status of patient health and take fast actions to improve patient condition. The use of smart sensors helps monitor different patient symptoms, recognize common disease patterns on the go and help doctor’s in quick diagnosis of patient health, especially important for mental healthcare [19], as mental health disorders are difficult to classify. People suffering from

Trends in Mental Health Based on IoT  239 Wearable device/sensor

Collect Data

Server/Cloud

Context Processing: Problem Detected

No

Continue Monitoring

Yes

Generate Alarm

Alert Family

Alert Doctor

Figure 12.3  IoT for mental health: A mechanism to alert caregiver.

mental health issues are increasing day by day; more than 21 million people are fighting with schizophrenia, about 35 million people are fighting with depression, and more than 47.5 million s people suffer from dementia [20]. Mental health and behavioral conditions of the patients have broad gap between each other so in that manner supplementary cost for people and society in general. Additionally, high economic costs for treatment have minimize the life hope and due to these disorders many years of patient’s life vanished because of mental, behavioral, and neurological diseases. Therefore, IoT is sustainable way out for the betterment of patient’s healthcare quality [18–20]. In [20], authors have explored four furthermost collective mental diseases to search out the sensors type which would identify exact symptoms to generate alarms regarding patient well-being. In this way, system can calculate the probability of which type of disease is affecting patient mental health by performing pattern matching of disease symptoms. In [21–23], authors have introduced the concept of using wearable technologies and smart sensing textile as a vital component of IoT ecologies, especially for the use of mental health well-being

240  The Smart Cyber Ecosystem for Sustainable Development monitoring. They proposed the protocol and conception of a data analysis prototype for the semantic analyze of the transcripts to recognize anxious and not so concerned participants. Based on results, there were substantial variances in between vocabulary of both concerned participants. This tactic offers an improved understanding of effects of smart wearable textiles and how it can help monitor patient reactions to different situations and environment conditions as part of observing a person’s mental well-being. In [24], to review the mental state behavior, the authors proposed a device which perceives some brain waves parameters like alpha, beta, and delta. This enables monitoring efficiently a person’s mental focus aka attention and observing breathing to analyze his/ her stress level. An important and very favorable procedure of reducing complexities was founded by this research and eHealth device is developed, which can be further improved with IoT applications. Hence, it is summarized that this system would give benefit to make health monitoring easier and in a cost efficient way. It showed tremendous potential with prototype sensing device, which helps observe patient well-being remotely in real time, hence help limit patient medical visits, helping in both reducing costs and patient stress levels [25]. The novel IoT technologies have potential to increase the delivery of quality services for patient health and hence enhance patient quality of life. IoT faces huge challenges as it tries to tackle huge quantity of data from connected devices, routing them to base stations and network formation for accomplishing consistent throughput, performance, and effectiveness, more important for smart mental healthcare [26, 27]. The classifications of psychiatric biomarkers contain genetics, proteins, and neuroimaging results. Other measures for monitoring a person’s mental health state and functioning are being discovered using smart technologies, which help in quick diagnosis of mental conditions, such as Alzheimer’s disease [28, 29]. Physical health also goes hand in hand with mental health, for example, Parkinson’s disease, which can be detected early using physiological health measures and decline in physical characteristics such as shaking and issues in balancing. At initial stage, detection of any mental health condition can benefit physicians to take early suitable actions, help control the advancement of the disease, and enhance the patient quality of life and reduce cost of healthcare [30]. In [29–31], the authors propose an application of IoT which contains of a model of activities performed by a person, like sitting, walking, carrying loads, cycling, exercising, and using different sensors to monitor personal health. In [31], people’s activity of gardening is used for monitoring people’s mental health. This application benefits the welfare and expands the mental health in elderly. In [32], Kim and Chung have proposed an index that measures depression by aggregating knowledge using a crowd source based application that utilizes the IoT, with an important aspect of the use of context information. Authors in [33] have proposed a system that observes signs for nightmares, and helps patient wake up slowly in order avoid unnecessary stress. In case patient does not wake up, it helps in post nightmare stress and improve patient mental well-being. The proposed system has the potential to save people’s lives and reduce depression with the help of IoT. There are many benefits of using IoT for mental health application, such as reduction in the treatment cost, reduction in human error, removal of geographical barrier, less paper work, chronic disorders can be detected early, better drug management for the patients (specially for elderly), alert used for medical attention, and improvement treatment results. These benefits are discussed in the following section.

Trends in Mental Health Based on IoT  241

12.5 Benefits of IoT in Mental Health 12.5.1 Reduction in Treatment Cost Mental well-being using IoT helps in reduction in the costs for both patient and healthcare services as IoT enables real-time monitoring of patients, with reduced patient visits to healthcare facilities and their remote monitoring by doctor’s. Doctors can also communicate via video and monitor different health parameters with the support of generation of alarms for critical situation. Moreover, it will help minimize insurance costs and unnecessary testing costs, as patient will only visit when necessary, as sensor’s can relay to them important information regarding blood pressure and sugar etc. using IoT [34].

12.5.2 Reduce Human Error IoT-based mental healthcare monitoring applications helps in minimizing human errors by monitoring vital information regarding blood pressure, blood sugar, heartbeat, etc., as sensor helps measure these parameters live, recognizes anomalies quickly quick data analysis, and takes immediate actions when necessary [35].

12.5.3 Remove Geographical Barriers Physicians and patients are connected world widely by internet so doctors can connect with patient from different geographical location and provide medical guidance to patients located at any place in the world [36]. This helps immensely, as one of the factors affecting the early detection and treatment is non-availability of the mental health experts. Especially, in the developing countries where number of doctors per patients are already low. This will help most importantly the children with autism, which if detected early and treated correctly then improve the quality of life for that child and also for his/ her parents.

12.5.4 Less Paperwork and Documentation IoT-based mental healthcare monitoring system improves coordination between stakeholders, such as nurses and doctors, by generating shared alarms and informing up-to-date patient information which will otherwise involve lengthy paperwork and documentation [37]. For example, sensors will give live information regarding vital diagnostic parameters: blood pressure, blood oxygen saturation, etc.

12.5.5 Early Stage Detection of Chronic Disorders Storing the IoT-based sensor information from patients and then utilizing big data analysis techniques chronic disorders can be found out early and treatment can begin early before disease prognosis becomes incurable [38].

242  The Smart Cyber Ecosystem for Sustainable Development

12.5.6 Improved Drug Management IoT-based drug delivery not only helps in making sure right medicine is reaching right patient; it can also help in inventory management easier. The supply chain management can also benefit, for example, using RFID (radio frequency Identification) technology to identify the depletion of a drug/medicine and trigger the need for its manufacturing and supply. Ultimately, it helps in reducing the costs by mismanagement, theft, and lost stock (detecting the expired stock) [39].

12.5.7 Speedy Medical Attention One of the primary IoT advantage is quick detection of emergency situation, for example, loss of blood oxygen saturation, by generating relevant alarms and alerting the staff and family to the patient needs [40]. Quick detection is very much important for the mental health of a child, for example, an autistic child, and also for the elderly, for example, early detection of dementia.

12.5.8 Reliable Results of Treatment One of the important advantage of IoT can be providing reliable and vital patient health parameters, using sensors, to the healthcare practitioner. It will help them make quick and better judgement regarding patient condition and improving patient mental health outcome in a timely manner [41].

12.6 Challenges in IoT-Based Mental Healthcare Applications 12.6.1 Scalability In markets, mental healthcare device would came to improve the quality of health in coming days. Very huge amount of devices connected to IoT on network ultimately would generate large amount of health data. Collection and process of the huge amount of data would also rise. This would create a big data issue for healthcare providers. Therefore, it is necessary to make the systems scalable as more and more IoT-based devices will be connecting to the system, generating large amount of data. Also, because the need for data storage and data analysis will be there as IoT-based devices generate data actively. This will entail a need for gaining knowledge and creating insight from that data for better mental health well-being [42].

12.6.2 Trust Trust is an important issue, in the usage of IoT devices and then utilizing the data for making decisions for proper mental health well-being. Principal issue is that of privacy as the patient data is most private and should only be visible and available to relevant healthcare practitioner. The patient information should be protected from attacks from third party for exploitation and should ensure its correctness at all time as important basis of mental health treatment plan. Also, reliability of data produced from the sensors from

Trends in Mental Health Based on IoT  243 their calibration to active life is an important trust issue. So, before IoT-based technology is deployed at large scale, the development of trust is a vital challenge and needs addressing on priority basis [43].

12.6.3 Security and Privacy Issues Healthcare devices and applications are associated to internet it captures patient’s private medical information. Therefore, it may attract hacker to steal medical data of patients. Medical information of any patient must be used by the authorization. The following are primary areas of concerns from security perspective in mental healthcare sector [44]: • • • •

IoT-based devices security Using protocol that transmit data securely Offer data security transparency, especially, in cloud computing environment By providing reduced consumption of resources, especially, the power, and maximum security for IoT devices • Providing a legal frame work for data protection and usage Blockchain-based solutions are increasingly becoming solution for security and reliable data transmission to the healthcare facilities. It will be discussed separately and considering its advantage whole new body of knowledge is focusing on blockchain in IoT.

12.6.4 Interoperability Issues As the IoT devices will be manufactured by different parties, vendors, etc., and connecting the devices with each other or with main system entails an inherent interoperability issue. It will raise the need for common interfaces and standardized data formats, as the data generated by different sensors can be stored easily and analyzed efficiently. Also important are standardized communication protocols, making sure that multiple devices can actually talk to each other or with the IoT-based healthcare system. Without addressing the issue of interoperability, the IoT-based generalized mental healthcare system will be a distant dream and use its potential [45].

12.6.5 Computational Limits One of the important issue affecting wide-scale deployment of IoT-based devices is the power consumption and consequently the devices use low power chips. This reduces their computation capability and limits their ability to perform intensive mental health related operations. This is a challenging issue and is being actively investigated in research, especially the use of energy efficient computation with secure protocols to transmit data [46].

12.6.6 Memory Limitations Medical devices in IoT healthcare have low memory. These types of devices are initiated through fixed OS (operating system). Thus, the memory of these devices is limited and not appropriate for performing the intricate routing and security protocols [47].

244  The Smart Cyber Ecosystem for Sustainable Development

12.6.7 Communications Media The IoT healthcare devices can be linked to the local network as well as the internet by the wide-ranging wireless connections like Zigbee, Bluetooth, GSM, or wirelessly. These entire wireless channels made old-style wired device’s security schemes less suitable. Hence, there is a need for a complete protocol which can handle both wired and wireless communication securely is a challenging task in itself [48].

12.6.8 Devices Multiplicity In an IoT healthcare environment, healthcare devices are assorted, ranging from PCs to RFID tags. The computation capability of these devices changes according to their computation, power, memory, and software. Hence, the challenge is to manipulate effective schemes that can accommodate the humblest of devices [49].

12.6.9 Standardization As new vendors join the IoT devices manufacturing race in healthcare sector, the issue of standardization rises. Which also creates issue of the interoperability. There is a need of standardization and regulation to govern the manufacturers. Especially important is setting of protocols, communications layers containing physical and MAC (media access control) layers, and device and gateway communication rules. Also, there is a need to store the information generated in a standardized manner to be utilized by different systems in healthcare. Many healthcare organizations and IoT research scholars work collectively to form the IoT mental healthcare technologies those are working to set the rules and regulation in healthcare sector based on IoT services [50, 51].

12.6.10 IoT-Based Healthcare Platforms The IoT-based healthcare requires development of new system architecture and design of platform that can function in low computation capability environments, with added requirements of transmission of data in a secure and timely manner. So, it requires new set of libraries and APIs. To construct the appropriate platform, a SOA (service-oriented approach) can be one of the example architecture for IoT-based mental healthcare system. Moreover, software developers and designers need to create well-organized development practices and templates for healthcare systems [52].

12.6.11 Network Type In the perspective of designing, an IoT mental healthcare networks can be in three basically diverse types: data centric, service centric, and design focusing foremost on patient. The approach based on data focuses on the generation of data and type of data being gathered from patients. Whereas, the service centric architecture focuses on the set of features being offered, and finally, patient centric architecture centers on the patient well-being and supporting their family members for making right treatment plan available [53].

Trends in Mental Health Based on IoT  245

12.6.12 Quality of Service Quality of service (QoS) is an important feature for any network and IoT is no exception. As healthcare, in general, and mental healthcare, in particular, involve real-time sensitive patient information, the need for guaranteed quality of service arises. It is an important challenge for IoT-based mental healthcare system to have QoS feature available as doctors need make right decisions in right time [54].

12.7 Blockchain in IoT for Healthcare As discussed, a critical challenge for healthcare system connected with IoT devices and other systems is the secure and reliable transmission of the data. Also, important issue is the data fragmentation and trust to bridge different IT systems with healthcare system and pharmaceutical companies [55]. So, blockchain not only helps in security and reliable transmission but also helps in sharing sensitive information between organizations, with added benefit of creating transparency between caregivers, family, and patients and collaboration between different healthcare organizations. Some of the work employing blockchain technology in IoT are discussed next, from use of hyperledger for preserving patient privacy to the multimedia transfer using the blockchain framework. As suggested by [56], the use of blockchain will be fundamental for preserving the privacy in the IoT, as IoT by nature are distributed device and favor a decentralized framework for security which is supported by the blockchain. Their simple blockchain frame work for IoT is represented in Figure 12.4. Other benefits of blockchain include immutable log of patient history and worldwide access to patient record for healthcare providers. In [57], they used IoT-based technology for patient vital sign monitoring and system integrity was maintained using hyperledger fabric-based blockchain system. It helps make IoT-based system secure and immune to illegal patient history changes. A detailed framework based on hyperledge is shown in Figure 12.5. A hybrid framework based on blockchain for processing data related to multimedia, such as images, audio, and videos is given by [58].

Health Alert

Blockchain Network Health Service provider

Patient monitored in home

React to alert, interact with patient

Figure 12.4  A simple frame work for the blockchain in IoT figure from [55].

246  The Smart Cyber Ecosystem for Sustainable Development

Enroll Request Escort

Patient

Notification WebSocket

Doctor

Peer

Nurse ECG Sensor REST Server Sphygmomanometer Sensor

Air flow Sensor

http Transaction

IoT Gateway EMG Sensor

Vital Sign Data

Patient SPo2 Pulsemeter

Peer

IoT Server

Orderee

Peer

gRPC

Wi-Fi/Ethernet (CoAP protocol)

Peer

Body Temperature Sensor Glucometer Sensor

Hyperledger Fabric Network

IoT Resources

Figure 12.5  Healthcare IoT blockchain platform implementation from [56].

12.8 Results and Discussion Survey included 89 total papers from search database. In addition, 74 papers were nominated for survey. From 74 articles, 15 articles were omitted because they did not fit with criteria. Six papers were deliberated the critical factors have influence on adoption of IoTbased healthcare applications. Two papers have not clearly well-defined the proper methodology. Three papers were not argued on network and device issues. Two papers just offered the success factors toward the adoption of mental healthcare applications. Two papers were student papers not the review papers. Overall, 58 articles were included in final survey. The aim of this survey is to summarize the mental healthcare applications and their benefits. These applications are useful to handle mental well-being they offer information about self-monitoring and awareness regarding mental health. Findings of this review illustrate many challenges behind implementation of IoT-based mental healthcare services. Some are discussed above in the section of challenges. Findings of this review suggested some recommendations: firstly, the maximize the understanding of challenges being faced by IoTbased technology’s successful deployment for patient mental health well-being. Secondly, data visualization is very important when the healthcare system is handling with IoT based devices. Finally, awareness programs must be arranged for those who are suffering from mental illness. Also important is the reliable and secure data sharing between care givers, healthcare practitioners, and organizations. For this, a framework based on the blockchain technology was studied and their importance for the IoT success is underlined. Blockchain technology not only helps to reduce data fragmentation but also helps to inhibit tempering with sensitive patient data.

Trends in Mental Health Based on IoT  247

12.9 Limitations of the Survey This survey has its shortcomings and it does not take into account systematic approach toward IoT-based devices that should govern the mental health well-being. A separate work is needed to analyze devices and sensors important for IoT application in mental healthcare. Also, important is issue of mode communication for these device and how to make communications secure and free from different types of attacks. Moreover, processes associated with the improvement of the applications, and issues of trust in these application was not detailed. Also not discussed were problems correlated to the generation of patient data and their analysis, like, data values missing. Finally, problems correlated to long-term exposure to wearable IoT devices in patients were not discussed.

12.10 Conclusion IoT-based healthcare, in general, and mental healthcare applications in IoT, in particular, bring many benefits, though with the benefits they also bring many challenges for their wide adaptation. In this work, we discussed basic architectures underpinning the IoT in healthcare. Also, shown the main advantages of using IoT for the mental health well-being. For example, reduction in the treatment cost, reduction in human error, removal of geographical barrier, less paper work, chronic disorders can be detected early, better drug management for the patients, etc. Also, in this work, we surveyed the challenges faced for wide-scale IoT deployment, such as scalability, trust, security, reliability, and QoS, etc. A blockchain-based solution was analyzed and found to be useful for building trust in IoT and improving security for IoT-based applications that can help its wider acceptability. These challenges illustrate that there is still an important research gap in the IoT-based mental healthcare and they show that IoT is not the solution for the mental health but is part of the solution. Its wide-scale deployment can help support the increasing population of mental health patients, especially in the elderly and in the children with autism. Therefore, there is big need of secure deployment of mental healthcare applications with IoT and requires further research in addressing other challenges in order to enable its wider implementation and acceptability in patients, care givers, and in the organizations.

References 1. Kalnoor, G., Shama, P.S., Suhashini, B.V.V.N., Krishna, B.M., Emergence Technologies, Smart Integration, Challenges and Applications of Internet of Things (IoT)in Electrical, Communication and Software Engineering. Proceedings of International Conference on Internet of Things (IConIoT), 2017. 2. Sharma, V. and Tiwari, R., A review paper onIOT& It’s Smart Applications. Int. J. Sci. Eng. Technol. Res. (IJSETR), 5, 2, 472–476, 2016. 3. Budida, D.A.M. and Mangrulkar, R.S., Design and implementation of smart Healthcare system using IoT. Innovations in Information Embedded and Communication Systems (ICIIECS) 2017 International Conference, pp. 1–7, 2017.

248  The Smart Cyber Ecosystem for Sustainable Development 4. Bauer, M. et al., Smartphones in mental health: a critical review of background issues, current status and future concerns. Int. J. Bipolar Disord., 8, 2, 1–19, 2020. 5. Parker, L., Bero, L., Gillies, D., Raven, M., Grundy, Q., “hot potato” of mental health app regulation: a critical case study of the Australian policy arena”. Int J Health Policy Manag, 8, 2, 168–76, 2019. 6. World Health Organization, Mental Health ATLAS 2017, International Review of Psychiatry, http://apps.who.int/iris/bitstream/handle/10665/272735/9789241514019-eng.pdf. 7. Hassanalieragh, M., Page, A., Soyata, T., Sharma, G., Aktas, M., Mateos, G. et al., Health monitoring and management using internet-of-things (IoT) sensing with cloud-based processing: opportunities and challenges. Proc. 2015 IEEE Int. Conf. Serv. Comput. SCC, vol. 7, no. 4, pp. 285–292, 2015. 8. Hamrioui, S. and Lorenz, P., Efficient wireless mobile networks communications applied to e-health. IEEE Int. Conf. Communication, vol. 2, no 8, pp. 1–6, 2017. 9. Lipschitz, J., Miller, C.J., Hogan, T.P., Burdick, K.E., Lippin-Foster, R., Simon, S.R. et al., Adoption of mobile apps for depression and anxiety: cross-sectional survey study on patient interest and barriers to engagement. JMIR Ment. Health, 22, 12, 13–17, 2019. 10. Goodday, S.M. and Cipriani, A., Challenges in identifying behavioral markers of bipolar disorder through objective smartphone data. Aust. N Z J Psychiatry, 53, 16, 9–18, 2019. 11. Huckvale, K., Torous, J., Larsen, M.E., Assessment of the data sharing and privacy practices of smartphone apps for depression and smoking cessation. JAMA Netw. Open, 19, 24, 25–42, 2019. 12. Wan, J., Gu, X., Chen, L., Wang, J., Internet of things for ambient assisted living: Challenges and future opportunities. Int. Conf. Cyber-Enabled Distrib. Comput. Knowl. Discov., pp. 354–357, 2017. 13. Fleming, T., Bavin, L., Lucassen, M., Stasiak, K., Hopkins, S., Merry, S., Beyond the trial: systematic review of real-world uptake and engagement with digital self-help interventions for depression, low mood, or anxiety. J. Med. Internet Res., 2, 44, 120–132, 2018. 14. Medrano-Gil, de los Ríos Pérez, S., Fico, G., Montalvá Colomer, C., Cea Sáncez, G., Umpierrez, C. et al., Definition of technological solutions based on the internet of things and smart cities paradigms for active and healthy ageing through Co-creation. Wirel. Commun. Mob. Comput., 27, 11, 1–15, 2018. 15. Aledavood, T., Hoyos, A.M., Alakörkkö, T., Kaski, K., Saramäki, J., Isometsä, E. et al., Data collection for mental health studies through digital platforms: requirements and design of a prototype. JMIR Res. Protoc., 6, 6, 110–123, 2017. 16. Martins, O., Osifeko, I., Gerhard, P., Hancke, I., Adnan, M., Abu-Mahfouz, I., Artificial Intelligence Techniques for Cognitive Sensing in Future IoT: State-of-the-Art, Potentials, and Challenges. J. Sens. Actuator Networks, 9, 21, 1–31, 2020. 17. Habibzadeh, H. et al., A Survey of Healthcare Internet of Things (HIoT): A Clinical Perspective. IEEE Internet Things J., 7, 1, 53–71, 2020. 18. Murali Krishna, B. et al., Novel Solution to Improve Mental Health by Integrating Music and IoT with Neural Feedback. J. Comput. Inf. Syst., 15, 3, 234–239, 2019. 19. Sugiura, K., Mahomed, F., Saxena, S., Patel, V., An end to coercion: rights and decision-making in mental healthcare. Bull. World Health Organiz., 98, 1, 32–52, 2020. 20. Chiu, M., Amartey, A., Wang, X., Vigod, S., Kurdyak, P., Trends in objectively measured and perceived mental health and use of mental health services: a population-based study in Ontario, 2002–2014. CMAJ, vol. 192, pp. 329–337, 2020. 21. Achtyes, E.D., Ben-Zeev, D., Luo, Z., Mayle, H., Burke, B., Rotondi, A.J. et al., Off-hours use of a smartphone intervention to extend support for individuals with schizophrenia spectrum disorders recently discharged from a psychiatric hospital. Schizophr Res., 8, 33, 200–206, 2019.

Trends in Mental Health Based on IoT  249 22. Zois, D., Sequential decision – Making in healthcare IoT Real Time health monitoring, Treatments and Interventions. IEEE 3rd World Forum Internet Things, vol. 22, no. 6, pp. 24–29, 2016. 23. Wu, C. and Zhang, Y., Toward efficient transparent computing for IoT apps by on-chip kernel offload. IEEE Internet Things J., 6, 3, 4085–4097, Jun. 2019. 24. Sahu, S. and Sharma, A., Detecting brainwaves to evaluate mental health using LabVIEW and applications. Proc. IEEE Int. Conf. Emerg. Technol. Trends Comput. Commun. Electr. Eng. ICETT, vol. 33, no. 5, pp. 3–6, 2017. 25. Aledavood, T., Hoyos, A.M.T., Alakörkkö, T., Kaski, K., Saramäki, J., Isometsä, E. et al., Data collection for mental health studies through digital platforms: requirements and design of a prototype. JMIR Res. Protoc., 6, 6, 1–10, 2016. 26. Alraja, M.N., Farooque, M.M.J., Khashab, B., The effect of security, privacy, familiarity, and trust on users’ attitudes toward the use of the IoT-based healthcare: The mediation role of risk perception. IEEE Access, 7, 6, 111341–111354, 2019. 27. Hayati, N. and Suryanegara, M., The IoT LoRa system design for tracking and monitoring patient with mental disorder. 2017 IEEE. Int. Conf. Commun. Netw. Satell. COMNETSAT 2017 – Proc, vol. 19, no. 44, pp. 135–139, 2017. 28. Bone, D., Lee, C., Chaspari, T., Gibson, J., Narayanan, S., Signal processing and machine learning for mental health research and clinical applications [perspectives]. IEEE Sign. Process. Mag., 34, 5, 196–195, 2017. 29. Lekjaroen, K., Ponganantayotin, R., Charoenrat, A., Funilkul, S., Supasitthimethee, U., Triyason, T., IoT planting:Watering system using mobile application for the elderly. 20th Int. Comput. Sci. Eng. Conf. ICSEC, vol. 3, no. 7, pp. 1–6, 2016. 30. Cosma, G., Brown, D., Battersby, S., Kettley, S., Kettley, R., Analysis of multimodal data obtained from users of smart textilesdesigned for mental wellbeing. Int. Conf. Internet Things Glob. Commun., vol. 10, no. 22, pp. 1–6, 2018. 31. Babar, M., Rahman, A., Arif, F., Jeon, G., Energy-harvesting based on internet of things and big data analytics for smart health monitoring. Sustain. Comput. Inform. Sys., 4, 53, 67–77, 2017. 32. Kim, C. and Chung, K., Depression index service using knowledge based crowdsourcing in smart health. Wirel. Pers. Commun., 9, 1, 255–268, 2017. 33. McWhorter, J., Brown, L., Khansa, L., Awearable health monitoring system for posttraumatic stress disorder. Biol. Inspired Cogn. Archit., 22, 23, 44–50, 2017. 34. Kajornkasirat, S. et al., Smart Health Monitoring System with IoT. IEEE, 8, 18, 206–211, 2018. 35. Mighali, V., Patrono, L., Stefanizzi, M.L., Rodrigues, J., Solic, P., A smart remote elderly monitoring system based on IoT technologies. Int. Conf. Ubiquitous Futur. Netw. ICUFN, vol. 2, no. 33, pp. 43–48, 2017. 36. Butkus, R., Rapp, K., Cooney, T., Engel, L.S., Envisioning a Better US Healthcare System for All: Reducing Barriers to Care and Addressing Social Determinants of Health. Ann. Internal Med., 172, 2, 50–59, 2020. 37. Jones, S.A., Azam, B., Morgan, K., Ahluwalia, N., An Audit of Documentation relating to Decision-Making Capacity at an Old Age Psychiatric Hospital in England. Int. J. Mental Health Capacity Law, 2019, 25, 49–60, 2020, 2020. 38. Hendrickx, D., Roeck, D., Maras, A., Dieleman, G., Gerritsen, S., Purper-Ouakil, D., Russet, F., Schepker, R., Signorini, G., Singh, S.P., Street, C., Challenges during the transition from child and adolescent mental health services to adult mental health services. BJPsych Bull., 45, 5, 1–6, 2020. 39. Bao, Y., Sun, Y., Meng, S., Shi, J., Lu, L., 2019-nCoV epidemic: address mental healthcare to empower society. Lancet, 3, 95, 102–124, 2020.

250  The Smart Cyber Ecosystem for Sustainable Development 40. Chen, Q. et al., Mental healthcare for medical staff in China during the COVID-19 outbreak. Lancet Psychiatry, 7, 4, 15–26, 2020. 41. Singh, P., Internet Of Things Based Health Monitoring System : Opportunities And Challenges. Int. J. Adv. Res. Comput. Sci., 9, 1, 224–228, 2018. 42. Milat, A. et al., Intervention Scalability Assessment Tool: A decision support tool for health policy makers and implementers. Health Res. Policy Syst., 18, 1, 33–45, 2020. 43. Miranda, J., Snowden, L., Legha, R., Policy Effects on Mental Health Status and Mental Healthcare Disparities, in: The Palgrave Handbook of American Mental Health Policy, vol. 15, no. 55, pp. 331–364, 2020. 44. Sogomonjan, M., Challenges and opportunities for e-mental health policy: an Estonian case study. Contemp. Social Sci., 4, 22, 1–14, 2020. 45. Isabel, D.L. et al., IoT-Based Services and Applications for Mental Health in the Literature. J. Med. Syst., 43, 11, 88–102, 2019. 46. Graham, S., Depp, C., Lee, E., Nebeker, C., Tu, X., Kim, H.C., Jeste, D., Artificial intelligence for mental health and mental illnesses: an overview. Curr. Psychiatry Rep., 21, 11, 1–16, 2019. 47. Gulliver, A., Farrer, L., Bennett, K., Griffiths, K.M., University staff mental health literacy, stigma and their experience of students with mental health problems. J. Further Higher Educ., 43, 3, 434–442, 2019. 48. Naslund, J., Aschbrenner, K., McHugo, G., Unützer, J., Marsch, L., Bartels, S., Exploring opportunities to support mental healthcare using social media: A survey of social media users with mental illness. Early Intervention Psychiatry, 13, 3, 405–413, 2019. 49. Ong, B.N., Hodgson, D., Small, N., Nahar, P., Sanders, C., Implementing a digital patient feedback system: an analysis using normalisation process theory. BMC Health Serv. Res., 20, 32, 1–16, 2020. 50. Ortega, P., Diamond, L., Alemán, M., Fatás-Cabeza, J., Magaña, D., Pazo, D., Pérez, V., Girotti, J., Ríos, E., Medical Spanish standardization in US Medical Schools: consensus statement from a multidisciplinary expert panel. Acad. Med., 95, 1, 22–31, 2020. 51. Selvaraj, S. and Sundaravaradhan, S., Challenges and opportunities in IoT healthcare systems: a systematic review. SN Appl. Sci., 2, 1, 1–39, 2020. 52. Kikuzawa, S., Pescosolido, B., Kasahara-Kiritani, M., Matoba, T., Yamaki, C., Sugiyama, K., Mental healthcare and the cultural toolboxes of the present-day Japanese population: Examining suggested patterns of care and their correlates. Social Sci. Med., 2, 28, 252–261, 2019. 53. McGraw, K., Adler, J., Andersen, S.B., Bailey, S., Bennett, C., Blasko, K., Greenberg, A., Hodson, S., Pittman, D., Ruscio, A., Mental healthcare for service members and their families across the globe. Military Med., 1, 84, 418–425, 2019. 54. Rushabh, S. and Alina, C., Iot and Ai In Healthcare: A Systematic Literature Review. Issues Inf. Syst., 19, 3, 33–41, 2018. 55. Dang, L.M., Piran, M.J., Han, D., Min, K., Moon, H., A Survey on Internet of Things and Cloud Computing for Healthcare. Electronics, 8, 7, 768, 2019. https://doi.org/10.3390/ electronics8070768. 56. Dwivedi, A.D., Srivastava, G., Dhar, S., Singh, R.A., Decentralized Privacy-Preserving Healthcare Blockchain for IoT. Sensors, 19, 326, 2019. 57. Jamil, F., Ahmad, S., Iqbal, N., Kim, D.-H., Towards a Remote Monitoring of Patient Vital Signs Based on IoT-Based Blockchain Integrity Management Platforms in Smart Hospitals. Sensors, 20, 2195, 2020. 58. Rathee, G., Sharma, A., Saini, H., Kumar, R., Iqbal, R., A hybrid framework for multimedia data processing in IoT-healthcare using blockchain technology. Multimedia Tools Appl., 79, 3, 1–23, 2020.

13 Monitoring Technologies for Precision Health Rehab A. Rayan1* and Imran Zafar2 Department of Epidemiology, High Institute of Public Health, Alexandria University, Alexandria, Egypt 2 Department of Bioinformatics and Computational Biology, Virtual University of Pakistan, Lahore, Pakistan 1

Abstract

Conventional health systems are aiming to diagnose first and treat diseases afterward. The growing precision health seeks early diagnosing and customized therapy for a health condition, by ongoing tracking of individual health and well-being. Portable and implantable devices can dynamically track and evaluate health, supplying a broad view of human living and health. Growing techniques are promising in powering the effect of precision health. For instance, the emerging monitoring techniques would continuously gather data and engage the customer. Novel techniques are steadily developing, hence the need for improved insights into the scenery of the current and future techniques which could gather data for precision health. This chapter outlines the available and growing tracking and sensing techniques for precision health, focusing especially on the highly needed techniques like portable and mobile devices, implantable sensors, and wearables. Keywords:  Precision health, health monitoring, sensors, wearables, healthcare

13.1 Introduction Precision health is an evolving technique for managing health. It promotes early disease diagnosing and prevention for subjects via counting for genetic, familial, living, and ecosystem differences. Precision health offers healthcare professionals and researchers a tool for expecting a disease onset and the highly efficient approaches for prevention and therapy on a personal level [1]. Precision health is a customized technique with a proactive rather than reactive healthcare. Modern innovations in technology and bio-sciences, such as epigenetics, gene sequencing, or micro-biome, could enable precision health in promoting accessible and efficient health services. Evaluating health and well-being dynamically could be enhanced by monitoring personal health via ongoing and sustainable detecting, collecting, and investigating data, involving: biological (like inflammation mediators), physical (like blood pressure), mental (like temper), behavioral (like exercising), and ecological (like contamination). Sensors *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (251–260) © 2021 Scrivener Publishing LLC

251

252  The Smart Cyber Ecosystem for Sustainable Development could be implantable devices, wearables, eatables, or fixed at home in someplace [1]. For instance, wearables could assess blood glucose level, body temperature, and heart rate [2]; a toothbrush could analyze saliva biologically [3]; a toilet could examine urine or feces digitally; a mirror could apply techniques like infrared or radar to assess vital signs [4]; a bed could monitor humidity, temperature, and cardiopulmonary performance [5]. Meanwhile, day-to-day activities, sleeping habits, or using a mobile phone could be related to tiredness, anxiousness, temper, or initial depression signals; and alarm the consumer, a relative, a friend, or a healthcare provider for dynamically handling the situation [6]. It is difficult to recognize the adequately valid, workable, and efficient technology among the latest and growing techniques (enormous kinds and quantities) for precision health which gathers and clusters valuable genuine data. Thorough tracking of potential novel techniques could considerably affect the economic and social growth in the long run. Multiple pieces of research unrevealed technical patterns to explain this technological leap [7]. For instance, concerning technological life cycle, technologies could be sorted as pre-emerging, emerging, or post-emerging [8]. Pre-emerging technologies are potential applications where technology holds both significant innovations and doubt along with limited consistency, development, or even a notable effect. Emerging technologies are promising novel applications offering solutions to a certain discipline where technology holds both significant innovations and doubt along with fair consistency, rapid development, and a notable effect, hence, inspiring remarkable societal and economic influence. Ultimately, post-emerging technologies are products or applications which might have been slightly presented to the market and hold both limited innovations and doubts. Universally, technology mega-trends are consistent market patterns to identify upcoming business, economic, and societal insights [4]. However, disruptive technologies could entirely transform the market’s landscape through dramatically changing business operations, institutions, or societies. Despite the tangled emergent and disruptive technologies, the shift from the latter to the former is not fully guaranteed and needs time [7]. Several research centers and agencies have evaluated technical trends, however, they lack investigating more in precision health. This chapter shows recent trends, potentials, limitations, and future insights of existing and emerging techniques optimum for precision health, focusing especially on sensing and monitoring devices. Such techniques provide decent innovation, homogeneity, and comparatively less doubt and are highly promising for invading the marketplaces.

13.2 Applications of Monitoring Technologies Emerging technologies in healthcare cover nanotechnology, medical image devices such as X-ray, MRI, or CT, 3D printing, virtual reality (VR) or augmented reality (AR), wearables, implantable biosensors, robotics, artificial intelligence (AI), machine learning (ML), cloud computing, Internet of Things (IoT), blockchain, and smart home sensors [9, 10]. In precision health, using smart biosensors and digital devices to monitor clinically related parameters such as biological, biochemical, or physiological indicators is among the leading techniques along with cloud computing and standards for interoperability [1]. Late research investigated health-related events applying innovative sensors that could monitor respiration, temperature or motion of the body, the rate of the heart, cognitive functions, and

Monitoring Technologies for Precision Health  253 blood parameters such as pressure, glucose, oxygen saturation, and pulse. Famous applications in health involve everyday life activities, well-being, patterns of breathing, calorie counters, and patterns of sleep.

13.2.1 Everyday Life Activities Ongoing monitoring of everyday life activities remotely could discover unusual conditions like falling accidentally [9]. Wearable sensors like accelerometers, magnetometers, or gyroscopes are getting common as life-assisting ambient technologies. By applying innovative algorithms to explore data, studies have investigated a hybrid technique comprising integrating smartphones’ gravity sensors with accelerometers to determine activity [2, 10]. Wrist accelerometers, magnetometers, and gyroscopes have been applied to determine the motion of the body and hence detecting falls [9, 11]. Likewise, Microsoft’s Kinect could detect falls by determining modifications in velocity, direction, and height [12]. A wearable wrist sensor attached to a seat-installed pressure detecting component was designed to identify and monitor in-seat activities [13]. Shoe-inserted sensors such as resistors for sensing force could detect daily life activities like laying down, sitting, standing, walking, driving, and house chores [14]. A silicon shoe insole model was developed integrating sensors for pressure and temperature and a gyroscope [15], while sensors for hetero-core optical load were developed to detect plantar pressure via monitoring gait, switches in natural weight and tumbling feet [16]. Many studies investigated sensors detecting eating activities where wrist gyroscopes and accelerometers could determine hand-based eating activity [17]. Monitoring motion of hands and breathing could expect the time period for a meal through wrist accelerometers and a specially designed belt over the chest and abdomen to enhance swallowing [18]. Tooth sensors could monitor consuming foods wirelessly via the mouth [19].

13.2.2 Sleeping and Stress Several devices have been developed to monitor sleeping and its quality. For instance, an ML-based wearable wristband was designed to monitor the level of sleepiness and the quality of sleeping via dermal temperature, movement, and heart rate [20]. Other models involve using a wrist android watch to monitor the quality of sleeping via heart rate sensors and accelerometers [21], devices evaluating mental and behavioral signs of dementia via monitoring stress and sleeping [22], and an intelligent pillow to identify sweating status and monitor the quality of sleeping via sensors for humidity and temperature [5]. Lately, psychological and physiological stress were found to affect one another. For instance, stress was predicted by applying accelerometers to measure the motions of the body and the rate of the heart and an intelligent wristband to measure body temperature [23]. The data from electrocardiogram (ECG), a biosensor for electrodermal activity (EDA), and a mobile electroencephalogram (EEG) headset was used to determine stress where it was associated with the cortisol levels in the saliva (the stress biomarker) [24]. An algorithm model was applied to monitor stress via biosensors for temperature, heart rate, and galvanic skin response (GSR), using smartphones to display data [25]. An ML algorithm-embedded EDA-wrist sensor is used to determine and identify stress and quiet behaviors, while another technique applied accelerometers, EDA, and sensors for dermal temperature

254  The Smart Cyber Ecosystem for Sustainable Development [26, 27]. Some researchers designed a glove-integrated sensor to assess the motion pattern of the steering wheel and hence predict a driver’s stress [28].

13.2.3 Breathing Patterns and Respiration Monitoring breathing patterns and respiration could yield important data about well-being, physical pressure, and burden during exercising. Prior studies investigated applying a wearable sensor for humidity produced by networks of light, adaptable, and extremely conducting permeable graphene to determine breathing and monitor respiration [29]. Another wearable sensor that is made of networks of magnetic resonance-adaptable intelligent fabric-based fiber Bragg was designed and verified to monitor respiration [30]. Other studies analyzed monitoring the activity of respiratory muscles using the electrical respiratory muscle activity of the diaphragm (EMGdi) and investigated a tiny light wireless sensor model to record the signals of EMGdi [31].

13.2.4 Energy and Caloric Consumption Quantifying energy and caloric consumption could enhance monitoring physical activity. For instance, a wrist device is particularly developed to monitor the elderly energy consumption and physical activity where it depends on accelerometers for movement, a pedometer, and heart rate to differentiate diverse levels of activity [21]. Other studies have emphasized especially on algorithms to forecast energy expenditure via wrist devices involving accelerometers, EDA, body temperature, electromyography (EMG), and sensors to monitor heart rate [32].

13.2.5 Diabetes, Cardiac, and Cognitive Care For diabetes, many models were designed to capture data on weight, blood pressure and glucose levels, and heart rate, and send these data to a smartphone for better visualization and monitoring conditions [33]. For heart health, a chest wearable to monitor physical and cardiac activity was developed where it involves ECG, accelerometers, sensors for temperature, a microcontroller, and a Bluetooth [34]. Other research estimated cardio-respiratory fitness via designing ML algorithms to explore data from heart rate, accelerometers, and Global Positioning System (GPS) [35]. For cognitive disorders, an intelligent ring with EDA, heart rate, accelerometers for movement, and sensors for dermal temperature was designed for monitoring the sympathetic nervous systems [36]. Other research investigated integrating accelerometers and force-sensitive resistors (FSRs) inside shoes to monitor both gait and activity in children suffering from cerebral palsy [37].

13.2.6 Disability and Rehabilitation In disability, for those who require a wheelchair, sensors-built wearable platform monitoring both EMG and heart rate along with a custom-packaged intrinsic monitoring system was designed allowing patients to monitor physical activity via a mobile app [38]. Some researchers investigated applying a sensor to identify and assess digitally deterioration in human mobility [39]. For rehabilitation, a wearable model was developed to monitor both

Monitoring Technologies for Precision Health  255 body motion and temperature and determine physical activity in patients on rehabilitations [40] where it comprises a wireless pulse oximeter and a movement-tracing device including a gyroscope, accelerometers, and an automatic movement processor. A textile-based stress sensor was developed to identify the angle of the knee joint [41].

13.2.7 Pregnancy and Post-Procedural Care In pregnancy, some studies investigated smart analysis of sensed blood pressure and physiological parameters such as proteinuria examinations to monitor the severity of hypertension in pregnancy and support medical decision-making [42]. For post-procedures, a wearable patch was designed to monitor respiratory and heart rates following surgeries in vulnerable patients where the patch is a medical, light, wireless, and wearable adhesive biosensor that could steadily assess several vital signs such as respiration rate, ECG, heart rate, dermal temperature, body posture, and steps [43].

13.3 Limitations This work reviewed the context of existing and growing automatic techniques for sensing and monitoring, which are promising for precision health, covering common sensors for measuring physiological parameters such as heart rate and EDA and body fluids such as blood glucose via implantable sensors or wearables such as wristbands, fabrics, and shoes. Such techniques emphasized diabetes, cardiovascular care, pregnancy, stress, and well-being. Yet, most of these techniques are tested at research levels, with limited commercially available options, especially for wearables. Designing, manufacturing, and adopting such novel techniques are both challenging and promising.

13.3.1 Quality of Data and Reliability There is a wide gap between research models and commercials. Market products usually depend on mature technologies, like recognized sensors, and affordable standards, while research models apply recent sensors with novel potentials, however still not grown to be included in market products [44]. According to the health application, research phase, or moral issues, selecting cautiously suitable sensors is a must. Most previously mentioned devices would likely be in the market or expected to be in the near future. The quality of the sensor’s gathered data directly affects analysis, and hence medical perspectives, however, collecting quality data is challenging. For example, using insufficiently calibrated or unstandardized sensors, changed behavioral and physiological parameters with no association to the monitored condition, for instance, movement and heart rate patterns, could seem irregular for a transit change in a patient’s living schedules rather than an underlying medical condition [45]. How precise and consistent the sensors are, is an issue, especially in marketed products. Novel technologies are not usually tagged as medical devices, and just limited numbers have proven effectiveness via clinical research [2]. Studies have revealed that data from the majority of marketed devices and sensors, such as activity trackers, is of lower precision and reliability compared to the physiological data collected via gold-standard measuring

256  The Smart Cyber Ecosystem for Sustainable Development devices. Hence, such devices need extensively verifying clinical studies for assessing efficiency and further enhancements [44]. Independent scientific verifications offer the highest available quality evidence in supporting such technologies. Yet, rigorous verifying standards for wearables and sensors could be challenging [2]. Technical improvements in battery life, developing sensors, and optimizing multi-sensors could gather data for verifications to mitigate precision issues [45].

13.3.2 Safety, Privacy, and Legal Concerns The safety specifications for software and hardware are challenging critically engineering and biomedicine since the human body is exposed to biosensors, wearables, or implantable devices [46]. Regulating bodies such as the FDA test health technologies for safety on marketing proposals, yet researchers claim a limited present role of such entities regarding mobile devices and wearables [45]. Besides, these entities do not govern purchasing or common use of automatic technologies such as wearables and smartphones, and mobile apps which accumulate personal health data and are not covered by their regulations. Hence, researchers, technology developers, and healthcare experts should pay more attention to solve the likely negative aspects and risks of these technologies. Accessing pervasive mobile technologies and IoT is growing; however, privacy and confidentiality should be tackled since there are social and ethical effects for the likely danger of inappropriately sharing such data [45].

13.4 Future Insights 13.4.1 Consolidating Frameworks Lately, there has been emerging research in establishing frameworks of wearables with consolidated sensors, like models of wireless body sensors [46]. Smart computing and devices have integrated wearables, mobile, and IoT devices to build consolidated frameworks for personal tracking and sensing, from physiological signs to physical activities, and daily interactions with the environment [47]. Ultimately, there would be a growing size, coupled with critically enhancing data reliability, of marketed wearables and implantable devices for monitoring of several physiological signs and activities such as heart rate, ECG, levels of blood glucose or pressure, EDA, motion, sites, among others.

13.4.2 Monitoring and Intervention With the growth of automatic technologies in health, passively monitoring and sensing are being implemented as an element of the healthcare process [45]. Recently, monitoring patients and intervening via smartphones, wearables, and IoT-centered techniques have been broadly investigated. Novel frameworks and sensors would augment research in fields like precision health for early detecting disorders or relapses, complying with therapies, and therapeutic effectiveness. Hence, soon, there would be a growing amount of marketed implantable devices or wearables such as patches for monitoring and intervening in blood glucose–related diseases such as diabetes or cardiac conditions. Incorporating

Monitoring Technologies for Precision Health  257 sensors-obtained data in the electronic health records helps to create a wider scope of a patient’s health. Healthcare professionals would use data from sensors, if sensors and other networking devices such as smartphones are technically inter-operable and consistent, and that is turning real via applying standards such as the Fast Healthcare Interoperability Resources (FHIR) [45].

13.4.3 Research and Development In research, portable frameworks have been supporting virtual clinical trials distantly such as home participants [45]. Growing wearables and mobile technologies would strengthen such trials via enhancing techniques for data acquisition, providing novel methods of collecting and analyzing physiological, activity, and clinical data to enable diagnosing and managing conditions. Patient’s self-reported data and passive sensing have grabbed researchers’ attention; yet, there is a gap between gathering data and achieving better patient results [45]. Using the likely unreliable patient’s self-reported data might be inconsistent for diagnosing diseases [2]. The growing application of integrated sensors necessitates to carefully choose sensors regarding various types for precisely gathering, visualizing, and intelligently analyzing raw data into valuable medical knowledge.

13.5 Conclusions This work reviewed emergent and post-emergent digital health and well-being technologies, especially highly demanding ones for sensing and monitoring data that are promising for precision health. The work highlighted technical trends, late advances, innovations, and applications of sensors in healthcare. Reaching the best results for precision health would require addressing technological and adoption limitations. This review could offer a broad scope and enriching perspectives on growing technologies, which might be applied in proactively managing personal health, and support customized and personalized decisions, monitoring, and therapy.

References 1. Gambhir, S.S., Ge, T.J., Vermesh, O., Spitler, R., Toward achieving precision health. Sci. Transl. Med., 10, 2018. https://doi.org/10.1126/scitranslmed.aao3612. 2. Peake, J.M., Kerr, G., Sullivan, J.P., A Critical Review of Consumer Wearables, Mobile Applications, and Equipment for Providing Biofeedback, Monitoring Stress, and Sleep in Physically Active Populations. Front Physiol, 9, 2018. https://doi.org/10.3389/fphys.2018.00743. 3. A.S, Dt, W., Salivary diagnostics: enhancing disease detection and making medicine better. Eur. J. Dent. Educ. Off J Assoc. Dent Educ. Eur., 12 Suppl 1, 22–9, 2008. https://doi. org/10.1111/j.1600-0579.2007.00477.x. 4. Sun, G., Negishi, T., Kirimoto, T., Matsui, T., Abe, S., Noncontact Monitoring of Vital Signs with RGB and Infrared Camera and Its Application to Screening of Potential Infection, in: NonInvasive Diagn Methods - Image Process, 2018, https://doi.org/10.5772/intechopen.80652. 5. Li, S. and Chiu, C., A Smart Pillow for Health Sensing System Based on Temperature and Humidity Sensors. Sensors, 18, 2018. https://doi.org/10.3390/s18113664.

258  The Smart Cyber Ecosystem for Sustainable Development 6. Sano, A., Phillips, A.J., Yu, A.Z., McHill, A.W., Taylor, S., Jaques, N. et al., Recognizing academic performance, sleep quality, stress level, and mental health using personality traits, wearable sensors and mobile phones. 2015 IEEE 12th Int. Conf. Wearable Implant. Body Sens. Netw. BSN, pp. 1–6, 2015, https://doi.org/10.1109/BSN.2015.7299420. 7. Mikova, N. and Sokolova, A.N., Global Technology Trends Monitoring: Theoretical Frameworks and Best Practices*, Foresight and STI Governance, 2014. 8. Joung, J. and Kim, K., Monitoring emerging technologies for technology planning using technical keyword based analysis from patent data. Technol. Forecast Soc. Change, 114, 281–92, 2017. https://doi.org/10.1016/j.techfore.2016.08.020. 9. Gjoreski, M., Gjoreski, H., Luštrek, M., Gams, M., How Accurately Can Your Wrist Device Recognize Daily Activities and Detect Falls? Sensors, 16, 2016. https://doi.org/10.3390/ s16060800. 10. Hassan, M.M., Uddin, MdZ, Mohamed, A., Almogren, A., A robust human activity recognition system using smartphone sensors and deep learning. Future Gener Comput Syst, 81, 307–13, 2018. https://doi.org/10.1016/j.future.2017.11.029. 11. Booth, R. and Goldsmith, P., A Wrist-Worn Piezoelectric Sensor Array for Gesture Input. J. Med. Biol. Eng., 38, 284–95, 2018. https://doi.org/10.1007/s40846-017-0303-8. 12. Nizam, Y., Mohd, M.N.H., Jamil, M.M.A., Development of a User-Adaptable Human Fall Detection Based on Fall Risk Levels Using Depth Sensor. Sensors, 18, 2018. https://doi. org/10.3390/s18072260. 13. Gravina, R. and Li, Q., Emotion-relevant activity recognition based on smart cushion using multi-sensor fusion. Inf. Fusion, 48, 1–10, 2019. https://doi.org/10.1016/j.inffus.2018.08.001. 14. Malvade, P.S., Joshi, A.K., Madhe, S.P., IoT based monitoring of foot pressure using FSR sensor. 2017 Int. Conf. Commun. Signal Process. ICCSP, pp. 0635–9, 2017, https://doi.org/10.1109/ ICCSP.2017.8286435. 15. Hubl, M., Pohl, O., Noack, V., Hahlweg, P., Ehm, C., Derleh, M. et al., Embedding of wearable electronics into smart sensor insole. 2016 IEEE 18th Electron. Packag. Technol. Conf. EPTC, pp. 597–601, 2016, https://doi.org/10.1109/EPTC.2016.7861550. 16. Otsuka, Y., Koyama, Y., Nishiyama, M., Watanabe, K., Extraction of natural weight shift and foot rolling in gait based on hetero-core optical fiber load sensor. Photonic Instrum. Eng. III, International Society for Optics and Photonics, vol. 9754, p. 975405, 2016, https://doi. org/10.1117/12.2212492. 17. Zhang, S., Stogin, W., Alshurafa, N., I sense overeating: Motif-based machine learning framework to detect overeating using wrist-worn sensing. Inf. Fusion, 41, 37–47, 2018. https://doi. org/10.1016/j.inffus.2017.08.003. 18. Dong, B. and Biswas, S., Meal-time and duration monitoring using wearable sensors. Biomed Signal Process Control, C, 97–109, 2017. https://doi.org/10.1016/j.bspc.2016.09.018. 19. Tseng, P., Napier, B., Garbarini, L., Kaplan, D.L., Omenetto, F.G., Functional, RF-Trilayer Sensors for Tooth-Mounted, Wireless Monitoring of the Oral Cavity and Food Consumption. Adv Mater Deerfield Beach Fla, 30, e1703257, 2018. https://doi.org/10.1002/ adma.201703257. 20. de Arriba-Pérez, F., Caeiro-Rodríguez, M., Santos-Gago, J.M., Collection and Processing of Data from Wrist Wearable Devices in Heterogeneous and Multiple-User Scenarios. Sensors, 16, 2016. https://doi.org/10.3390/s16091538. 21. Alfeo, A.L., Barsocchi, P., Cimino, M.G.C.A., La Rosa, D., Palumbo, F., Vaglini, G., Sleep behavior assessment via smartwatch and stigmergic receptive fields. Pers. Ubiquitous Comput., 22, 227–43, 2018. https://doi.org/10.1007/s00779-017-1038-9. 22. Kikhia, B., Stavropoulos, T.G., Meditskos, G., Kompatsiaris, I., Hallberg, J., Sävenstedt, S. et al., Utilizing ambient and wearable sensors to monitor sleep and stress for people with BPSD in

Monitoring Technologies for Precision Health  259 nursing homes. J. Ambient Intell. Humaniz Comput., 9, 261–73, 2018. https://doi.org/10.1007/ s12652-015-0331-6. 23. Yu, Y., Wang, Q., Liu, J., Pervasive health monitor and analysis based on multi-parameter smart armband. 2015 37th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBC, Milan: IEEE, pp. 5493–6, 2015, https://doi.org/10.1109/EMBC.2015.7319635. 24. Betti, S., Lova, R.M., Rovini, E., Acerbi, G., Santarelli, L., Cabiati, M. et al., Evaluation of an Integrated System of Wearable Physiological Sensors for Stress Monitoring in Working Environments by Using Biological Markers. IEEE Trans. Biomed. Eng., 65, 1748–58, 2018. https://doi.org/10.1109/TBME.2017.2764507. 25. Lebepe, F., Niezen, G., Hancke, G.P., Ramotsoela, T.D., Wearable stress monitoring system using multiple sensors. 2016 IEEE 14th Int. Conf. Ind. Inform. INDIN, pp. 895–8, 2016, https:// doi.org/10.1109/INDIN.2016.7819288. 26. Gjoreski, M., Luštrek, M., Gams, M., Gjoreski, H., Monitoring stress with a wrist device using context. J. Biomed. Inform., 73, 159–70, 2017. https://doi.org/10.1016/j.jbi.2017.08.006. 27. Zangróniz, R., Martínez-Rodrigo, A., Pastor, J.M., López, M.T., Fernández-Caballero, A., Electrodermal Activity Sensor for Classification of Calm/Distress Condition. Sensors, 17, 2017. https://doi.org/10.3390/s17102324. 28. Lee, B.-G. and Chung, W.-Y., Wearable Glove-Type Driver Stress Detection Using a Motion Sensor. IEEE Trans. Intell. Transp. Syst., 18, 1835–44, 2017. https://doi.org/10.1109/ TITS.2016.2617881. 29. Pang, Y., Jian, J., Tu, T., Yang, Z., Ling, J., Li, Y. et al., Wearable humidity sensor based on porous graphene network for respiration monitoring. Biosens Bioelectron, 116, 123–9, 2018. https:// doi.org/10.1016/j.bios.2018.05.038. 30. Massaroni, C., Saccomandi, P., Formica, D., Lo Presti, D., Caponero, M.A., Di Tomaso, G. et al., Design and Feasibility Assessment of a Magnetic Resonance-Compatible Smart Textile Based on Fiber Bragg Grating Sensors for Respiratory Monitoring. IEEE Sens J., 16, 8103–10, 2016. https://doi.org/10.1109/JSEN.2016.2606487. 31. Estrada, L., Torres, A., Sarlabous, L., Jané, R., Evaluating respiratory muscle activity using a wireless sensor platform. 2016 38th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBC, pp. 5769– 72, 2016, https://doi.org/10.1109/EMBC.2016.7592038. 32. Ingraham, K.A., Ferris, D.P., David Remy, C., Using wearable physiological sensors to predict energy expenditure. IEEE Int Conf Rehabil Robot Proc, vol. 2017, pp. 340–5, 2017, https://doi. org/10.1109/ICORR.2017.8009270. 33. Alfian, G., Syafrudin, M., Ijaz, M.F., Syaekhoni, M.A., Fitriyani, N.L., Rhee, J., A Personalized Healthcare Monitoring System for Diabetic Patients by Utilizing BLE-Based Sensors and RealTime Data Processing. Sensors, 18, 2018. https://doi.org/10.3390/s18072183. 34. Khan, J.A., Akbar, H.A., Pervaiz, U., Hassan, O., A wearable wireless sensor for cardiac monitoring. 2016 IEEE 13th Int. Conf. Wearable Implant. Body Sens. Netw. BSN, pp. 61–5, 2016, https://doi.org/10.1109/BSN.2016.7516233. 35. Altini, M., Casale, P., Penders, J., Amft, O., Cardiorespiratory fitness estimation in free-living using wearable sensors. Artif. Intell. Med., 68, 37–46, 2016. https://doi.org/10.1016/j. artmed.2016.02.002. 36. Mahmud, M.S., Fang, H., Wang, H., An Integrated Wearable Sensor for Unobtrusive Continuous Measurement of Autonomic Nervous System. IEEE Internet Things J., 6, 1104–13, 2019. https:// doi.org/10.1109/JIOT.2018.2868235. 37. Hegde, N., Zhang, T., Uswatte, G., Taub, E., Barman, J., McKay, S. et al., The Pediatric SmartShoe: Wearable Sensor System for Ambulatory Monitoring of Physical Activity and Gait. IEEE Trans. Neural Syst. Rehabil. Eng. Publ .IEEE Eng. Med. Biol. Soc., 26, 477–86, 2018. https:// doi.org/10.1109/TNSRE.2017.2786269.

260  The Smart Cyber Ecosystem for Sustainable Development 38. Pal, A., Monsalvo, K., Sunthonlap, J., Arguelles, P., Adame, A., Tu, J. et al., Wheelchair Exercise Monitor Development Platform - An Application for Wireless EMG Sensors. Proceedings of the 7th International Conference on Sensor Networks - Volume 1: SENSORNETS, Funchal, Madeira, Portugal, 2018, pp. 67–73, 2020. 39. Leightley, D., McPhee, J.S., Yap, M.H., Automated Analysis and Quantification of Human Mobility Using a Depth Sensor. IEEE J. Biomed. Health Inform., 21, 939–48, 2017. https://doi. org/10.1109/JBHI.2016.2558540. 40. Kańtoch, E., Human activity recognition for physical rehabilitation using wearable sensors fusion and artificial neural networks. 2017 Comput. Cardiol. CinC, pp. 1–4, 2017, https://doi. org/10.22489/CinC.2017.296–332. 41. Gholami, M., Ejupi, A., Rezaei, A., Ferrone, A., Menon, C., Estimation of Knee Joint Angle Using a Fabric-Based Strain Sensor and Machine Learning: A Preliminary Investigation. 2018 7th IEEE Int. Conf. Biomed. Robot. Biomechatronics Biorob, pp. 589–94, 2018, https://doi. org/10.1109/BIOROB.2018.8487199. 42. Moreira, M.W.L., Rodrigues, J.J.P.C., Oliveira, A.M.B., Saleem, K., Smart mobile system for pregnancy care using body sensors. 2016 Int. Conf. Sel. Top. Mob. Wirel. Netw. MoWNeT, pp. 1–4, 2016, https://doi.org/10.1109/MoWNet.2016.7496609. 43. Breteler, M.J.M., Huizinga, E., van Loon, K., Leenen, L.P.H., Dohmen, D.A.J., Kalkman, C.J. et al., Reliability of wireless monitoring using a wearable patch sensor in high-risk surgical patients at a step-down unit in the Netherlands: a clinical validation study. BMJ Open, 8, e020162, 2018. https://doi.org/10.1136/bmjopen-2017-020162. 44. Kamišalić, A., Fister, I., Turkanović, M., Karakatič, S., Sensors and Functionalities of NonInvasive Wrist-Wearable Devices: A Review. Sensors, 18, 1714, 2018. https://doi.org/10.3390/ s18061714. 45. Reinertsen, E. and Clifford, G.D., A review of physiological and behavioral monitoring with digital sensors for neuropsychiatric illnesses. Physiol. Meas, 39, 05TR01, 2018. https://doi. org/10.1088/1361-6579/aabf64. 46. Leu, F., Ko, C., You, I., Choo, K.-K.R., Ho, C.-L., A smartphone-based wearable sensors for monitoring real-time physiological data. Comput. Electr. Eng., 65, 376–92, 2018. https://doi. org/10.1016/j.compeleceng.2017.06.031. 47. Lee, U., Han, K., Cho, H., Chung, K.-M., Hong, H., Lee, S.-J. et al., Intelligent positive computing with mobile, wearable, and IoT devices: Literature review and research directions. Ad Hoc Netw., 83, 8–24, 2019. https://doi.org/10.1016/j.adhoc.2018.08.021.

14 Impact of Artificial Intelligence in Cardiovascular Disease Mir Khan1, Saleem Ahmed2*, Pardeep Kumar3 and Dost Muhammad Saqib Bhatti2 National Institute of Cardiovascular Diseases, Karachi, Pakistan Dawood University of Engineering & Technology, Karachi, Pakistan 3 Quaid e Awam University of Engineering, Science & Technology, Nawabshah, Pakistan 1

2

Abstract

The field of medicine has made tremendous progress; however, cardiovascular disease (CVD) is still the major reason of death cause in the present world. Still, major studies are required to improve CVD death ratio. One of the areas which can be exploited in order to reduce CVD is by working on medical imaging. The Artificial Intelligence (AI) strategies such as machine and deep learning with advance clinical data can be benefited to make the unavoidable healthcare benefit through which elderly and constant infection patients can get restorative care at their domestic, lessening hospitalizations and making strides in quality of life. The CVD may be a key chance to well-being and the critical source of death worldwide. The event of CVD created 17.6 million deaths in 2016, an increment of 14.5% from 2006 to 2016 [1]. The mortality and disease rates of CVD are growing yearly, specifically in emerging countries [2]. Reports have shown that up to around 80% of CVD-related deaths occur in developing countries. Other than that, these deaths occur at a very young age in developed countries [3]. Furthermore, current pandemic COVID-19 has also higher impact on CVD patients. CVD has put an overwhelming burden on patients and civilization as an entire. Hence, it is necessary to develop techniques for moving forward with determination to cure the CVD in coming years for better future. The AI can be one of the sources and can help to find the solutions for CVD related diseases. Keywords:  Artificial intelligence, medical imaging, machine learning, convolutional neural networks

14.1 Artificial Intelligence Artificial Intelligence (AI) is one of the field of computer science which is intelligence shown by machine rather than humans. The AI machines can understand human speech, playing games, autonomous cars, and many others. Similarly, AI can play important role in healthcare. Machine Learning (ML) can be employed to develop various algorithms by *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (261–272) © 2021 Scrivener Publishing LLC

261

262  The Smart Cyber Ecosystem for Sustainable Development using medical and healthcare data. The basic role of AI in healthcare is to analyze relationship between patients’ outcome and treatment methods. Another use of AI in healthcare can be used for diagnoses and to prevent the disease by applying suitable ML and deep learning algorithms. Besides this, AI can be used in pharmaceutical field which can have further impact on healthcare issues like cardiovascular disease (CVD) [4]. The rapid progress of AI methods, mostly in the subdomains of ML and deep learning has rapidly concerned the courtesy of clinicians to generate new unified, consistent, and effective approaches in order to solve healthcare issues. AI methods can also support physicians to improve medical conclusions empowering early discovery of subclinical organ failure, with clinical pertinence.

14.2 Machine Learning ML helps the system to make better predictions and helps making better decision by involving proper data that engages calculations to induce it and learn from it. In addition, the noteworthy of ML is to find the hidden patterns from information and create an outline based on that information. Along these lines, the machine can utilize this to provide future patterns [5]. ML could be a prevalent sub discipline of AI, which represents different methods for tackling complicated issues with huge information by distinguishing interaction designs among factors. As compared to conventional measurement approaches, ML is centered on building robotized clinical choice frameworks (such as read-mission and mortality score frameworks) that help doctors make more exact forecasts, instead of basic assessed score frameworks. ML can be categorized into following three categories as shown in Figure 14.1. Three categories of ML: 1. Supervised Learning: It is important in supervised learning to train the algorithm. There are different algorithm in supervised learning like regression and convolutional neural networks (CNNs). The supervised learning can be utilized in healthcare systems to predict or analyze various medical parameters based on the medical/patient records. 2. Unsupervised Learning: In this method, the data set is without labels; therefore, it suggests that the machine must discover the label itself. Unsupervised learning can play important role in CVD disease treatment and can support solving issues such as CVD forecast, cardiovascular image analysis, conclusion, and treatment [5]. 3. Reinforcement Learning: In reinforcement learning, there is no necessity to complete the offered objectives. In supervised and unsupervised learning, humans pose an objective and it should be accomplish by applying one of the mentioned method. Another approach of AI is based on response mechanism, frequently defined as a “reward”. The objective of this category of ML is not to reach the offered goal, but to exploit the reward for the model throughout the learning process. It has been deployed to enhance the process of patients who are on mechanical ventilation in intensive care units (ICUs).

Impact of AI in CVD  263 Machine Learning/Deep Learning

Supervised Learning

Classification Identifying to which of a set of categories a new observation belongs

Regression Predicting what value (continuous) a new observation take

Reinforcement Learning

Unsupervised Learning

Clustering Divide individuals into groups with similar characteristics

Others

Learn an action that achieves the best rewards for the agent in environment

Dimensionality reduction, anomaly detection, association rules, etc

Figure 14.1  Supervised, unsupervised, and reinforcement learning.

14.3 The Application of AI in CVD AI advances have been useful in cardiovascular medicine counting precision medicine, clinical expectation, cardiac imaging examination and intelligent robots, and cardiovascular medicine.

14.3.1 Precision Medicine Artificial intelligence can be essentially utilitarian for far off subsequent meet-ups, moment sicknesses guiding, ideal alerts of signs, and drug prompts. Simultaneously, from the impression of clinicians, artificial intelligence can assist assemble with voicing data, connect electronic clinical records frameworks, and lessening the remaining task at workload of clinicians [6]. Within the future, intellectual computers (gadgets are taught by machine or deep learning calculations and can unravel issues deprived of human help) will offer assistance to clinicians in order to make precise choices and forecast patient results. With the assistance of AI, it is most likely to execute a precise therapeutic that modifies healthcare for each individual. AI may not replace clinicians but clinicians can utilize AI innovations to innovate in CVD cure and drug development.

14.3.2 Clinical Prediction By using ML and huge information analytics, AI can offer assistance to clinicians to form forecasts that are more precise for patients. Research from Dawes TJW recommends that AI can foresee conceivable times of passing for heart illness patients [7]. In their research, AI program recorded the cardiac magnetic resonance imaging (MRI) and blood lab tests of 256 heart illness patients. The program measured the development of 30,000 meeting

264  The Smart Cyber Ecosystem for Sustainable Development points that are checked on the heart structures in each pulse. By merging this information with the patients’ 8 years health records, AI seem to foresee the irregular conditions that will lead to persistent passing. Moreover, their program was able to foresee the survival rates of patients for another 5 years, and another year of survival for patients having chances of 80%. Moreover, Motwani and his colleagues set up a prophetic prediction using deep learning, in order to increase 5-year life time, for 10,030 suspected coronary heart infection (CHD) patients. This study shows that the evaluation based on AI has higher accuracy compared to conventional medical judgment and coronary computed tomographic angiography [8].

14.3.3 Cardiac Imaging Analysis With the invention of ML, cardiac imaging examination has shown incredible improvement. The ML can offer assistance to analyze electrocardiogram (ECG), coronary angiography, and echocardiography. In later decades, the cardiac medication has most focus on the counting CHD and acute coronary syndrome (ACS). The ML algorithms can recognize coronary atherosclerotic more precisely than clinicians. In addition, AI can too be utilized to analyze echocardiographic pictures. The College of California, San Francisco, made convolutional neural systems through utilizing the echocardiographies of 267 randomized patients (age range: from 20 to 96 years) between 2000 and 2017 from the college therapeutic center. The 223,000 pictures were isolated into 15 groups. Moreover, this grouping calculation has outflanked the human cardiovascular doctors within the classification competition of cardiac ultrasound images. The deep learning will make imaging analysis more accurate at ease and early prediction before it becomes severe [9].

14.4 Future Prospect AI has made a lot of progress in cardiovascular medicine. The incorporation of AI and cardiovascular medicine involves qualified services and advanced technologies. The AI ventures are most likely to be conducted by huge innovation enterprises such as Google, Apple, and Microsoft which have contributed intensely in AI to progress the efficiency of cardiovascular medicine. Stanford and Apple propelled a venture entitled “Apple Heart Study” with the assistance of ML. In addition, the advancement of sensor innovation has encouraged the use of AI in cardiovascular medication. The latest Apple watch series 4 has a new transducer that measures ECG. The US Food and Drug Administration (FDA) have approved this new feature. In early 2018, researchers from Verily (Google Life Sciences-Alphabet Inc. inquire about organization) utilized ML to evaluate the hazard of sympathetic enduring from cardiovascular infection. They effectively performed study to analyze the patient’s eye. At that point, they also gathered different sorts of information, counting the patient’s blood, age, weight, and smoking status. Subsequently, this permitted the researchers to foresee the patient’s chance of cardiovascular infection. To prepare the calculation, they utilized ML to analyze the therapeutic information of about 30 million patients. As an outcome, the precision of the calculation in recognizing patients with cardiovascular illness was as high as 70%, which is near to the conventional cardiovascular treatment [10].

Impact of AI in CVD  265 Besides, Microsoft as of late reported that they will help out Apollo Medical clinic in India recorded as hard copy calculations to support clinicians in anticipating the danger causes for CVD. On 9 January 2017, the FDA offered freedom for the utilization of a cardiovascular X-ray examination programming called Cardio DL (from Arterys) that utilizes profound learning for clinical picture investigation and gives mechanized ventricular division to conventional heart X-ray checks. By utilizing distributed computing, Cardio DL can naturally finish picture preparing in under 10 seconds, and can draw the framework of the ventricular epicardium and subcardium, to precisely assess the capacity of the ventricle [11]. Siemens has constructed a huge information base of in excess of 250 million connected pictures, reports, careful information, and different materials for preparing its simulated intelligence computation programs. A team of cardiologists at the University Hospital of Heidelberg conducted a 6-year trial. A group of cardiologists at the University Hospital of Heidelberg led a 6-year preliminary. They utilized information from patients with cardiovascular breakdown to produce 100 carefully mimicked hearts and utilized artificial intelligence to foresee the forecast of these patients, and afterward contrasted the anticipated outcomes and the genuine circumstance of the patients. Clinicians can even utilize 3D production innovation to make models of the heart, to build up a more suitable cure. All these information recommend that a significant insurgency in the medical use of artificial intelligence in cardiovascular medication (closely resembling a Cambrian blast) may happen shortly, and this application is just the start of the general utilization of computer-based intelligence.

14.5 PUAI and Novel Medical Mode 14.5.1 Phenomenon of PUAI The problems clinical industry has experienced, the greater part of the current ramifications of computer-based intelligence in cardiovascular medication can be depicted as “PUAI”. In clinical practice, the principle issue looked by specialists consistently ought to be understood first, for example, the right analysis and powerful treatment for understanding. The application and abilities of computer-based intelligence depend for huge scope and develop clinical data for AI. At present, it very well may be applied in certain particular conditions, for example, trauma center and chest pain center (CPC). With the improvement of artificial intelligence innovation, we actually cannot ensure that the innovation is solid. A few specialists stressed that a few clinicians will depend altogether on computer-based intelligence to manage patients. Whereas, specialists are the pillar and artificial intelligence can assist specialists with improving the adequacy of their treatment [12]. Verghese et al. called attention to those clinicians that can utilize computerbased intelligence to more readily serve patients [13]. Studies have proposed that the blend of clinicians and artificial intelligence abilities will furnish patients with better symptomatic outcomes than experience alone [1]. To this end, in light of the fact that our group has developed another territorial agreeable salvage model to upgrade the conclusion and therapy framework for CPCs [14], we have planned another clinical model that may enable youthful clinicians to decrease the pace of misdiagnosis [15].

266  The Smart Cyber Ecosystem for Sustainable Development Traditional mode and Novel mode Instructions Clinician

Patient Traditional Mode

Novel Mode Instructions Clinician

Correct Instructions Al

Patient

Debatable Instructions Request help from senior clinician

Figure 14.2  The clinician can pass the finding plan through artificial intelligence.

14.5.2 Novel Medical Model Right now, clinicians give patients direct treatment choices dependent on their personal finding. Nonetheless, because of the clinician’s understanding, stress, or exhaust, different causes may make an off-base finding, and even lead to disastrous results. In the new model as shown in Figure 14.2, the clinician can pass the finding plan through artificial intelligence, and if the guidelines are right, the simulated intelligence will accomplish. In the event that the guidelines are scrambled, disregard the directions given by the artificial intelligence dependent on AI and approach the senior clinician for help. It is trusted that new model will decrease the rate of clinical misbehavior brought about by clinician blunders.

14.6 Traditional Mode 14.6.1 Novel Medical Mode Plus PUAI We can join the idea of PUAI into novel clinical model. We can locate a simpler method to fabricate an artificial intelligence model. Presently, the restricted utilization of computerbased intelligence for clinical conclusion includes the differential finding of CVD. There are a few kinds of CVD, each with various analytic strategies; along these lines, complex calculations and models, for example, profound learning and fortification learning, are needed to expand the trouble of composing calculations and increment the size of preparing informational indexes and computer-based intelligence models. Be that as it may, by fusing PUAI into another clinical model, it is just important to plan a little and basic model for the analysis of basic infections, for example, ACS and aortic dismemberment by utilizing less difficult calculations. There is no compelling reason to stress over the restrictions of

Impact of AI in CVD  267 flow artificial intelligence obstruction between these ailments. For instance, these kinds of thoughts can be applied to build up a notice framework in the current clinical framework, as shown in Figure 14.3. The patient arrives in the emergency clinic, after that, the patient’s data is all the while gone into the information base when the clinician gathers the history. In view of the clinician’s finding, the simulated intelligence will be agreeing the symptomatic models for the sickness (Standard A) from the information base, contrasted with the patient’s genuine condition (Standard B). In the event that the examination results coordinate, the admonition framework will not be ready. In case of clashing outcomes, the simulated intelligence will create an alarm, making the clinician aware by cautiously looking at his/ her finding. This new medical admonition framework is appropriate for ICUs, as well as the coronary consideration unit and CPC, which is particularly valuable during night shifts. Since, in these particular regions, the presentation of night shifts specialists directly affects persistent safety [16]. Studies from Maltese et al. have demonstrated that the dynamic capacity of ICU specialists has dropped altogether [17]. Presently, this new clinical admonition framework shows incredible potential for evading misdiagnosis brought about around evening time move clinicians’ psychological decrease. In addition, the notice framework is anything but difficult to apply. Our group has planned another provincial helpful salvage model to upgrade the analysis and treatment framework for CPCs. It gives ideal and powerful PCI to patients with ST-fragment height myocardial localized necrosis, particularly in creating nations, for example, China [5]. The central issue of this model is to lessen the time from indication beginning to reperfusion and cardiovascular mortality. Later on, we will likely apply this novel clinical model to existing frameworks. By utilizing an AI-based calculation, artificial intelligence can cautiously inspect the clinician’s determination and cure strategy. The objective is to save more lives. We will keep on refining this novel clinical model and check its functional application esteem in medical practice.

Clinician

Database

Diagnosis

1

Artificial Intelligence 1

Clinical Manifestation

2

Patient’s Information

2

Standard A

Standard B

Clinical Manifestation Physical Examination

Physical Examination Imaging Examination

Imaging Examination

Laboratory Examination

Laboratory Examination Other factors

Other factors

Comparison

Engaged

Warning System Mismatch

Figure 14.3  Current clinical framework.

Disengaged Match

268  The Smart Cyber Ecosystem for Sustainable Development At present, CVD stays a significant medical issue influencing the whole world, particularly in low and center salary nations. It will keep on being the primary driver of mortality in the next 20 years. The utilization of simulated intelligence, particularly AI, has demonstrated incredible potential in overseeing and treating this irksome illness [3, 18]. In addition, we solidly accept that later on, computer-based intelligence will be a collaborator to clinicians, not a foe, since computer-based intelligence was initially intended to mirror human reasoning cycles as opposed to advancement. Hence, the clinician ought to comprehend the importance of simulated intelligence and be acquainted with its use esteem. The capacity of computer-based intelligence will increment as clinicians develop further mindfulness of the infection. The improvement of clinical abilities, inside, and out clinical exploration is the reason for the advancement of artificial intelligence. Clinicians must not disregard to persistently figure out how to recover their capacity to help patients, and not to depend a lot on machines and simulated intelligence.

14.7 Representative Calculations of AI ML and deep learning consist of a multitude of algorithms. Table 14.1 summarizes brief descriptions of basic ML algorithms used in different tasks. Currently, ensemble learning and deep learning can be described as the mainstay of algorithms of AI. Ensemble learning is a ML method that combines multiple “weak” learners (algorithms) such as decision tree and logistic regression (Table 14.1) to obtain a good prediction. Boosting, bagging, and stacking are the three main methods of ensemble learning [19].

14.8 Overview of Pipeline for Image-Based Machine Learning Diagnosis The in general pipeline to construct ML instruments for image-based cardiac determination is schematically portrayed within the taking after segment, as well as in Figure 14.4. In brief, it requires (1) input imaging datasets from which appropriate imaging indicators can be extricated, (2) precise yield determination names, and (3) a suitable ML procedure that’s ordinarily chosen and optimized depending on the application to foresee the cardiac determination (yield) based on the imaging indicators (input). Extra non-imaging indicators (e.g., ECG information, hereditary information, sex, or age) regularly coordinate to demonstrate the ML and regularly move forward to demonstrate the execution. In this segment, we will begin to talk about the input and yield factors in more details and, sometime recently, present common utilized ML procedures and their applications.

Impact of AI in CVD  269 Table 14.1  Brief descriptions of basic machine learning algorithms used in different tasks. Algorithm

Description

Use

Logistic regression

An algorithm that estimates probability of dichotomized outcome from multiple covariates using logistic function.

Classification

Decision tree

A flow chart–like algorithm that divides data into branches by considering information gain. The final branches represent output of the algorithm (class or value).

Classification/regression

(simple) Neural network

An algorithm inspired by human brain architecture. Layers consisting of nodes are connected to one another with edges weighted as per training results.

Classification/regression

K nearest neighbor

A simple algorithm that classifies observations by comparing k examples that exist in the nearest locations (= examples with the most similar features).

Classification/regression

Support vector machine

Support vector machine draws a boundary line that maximizes margins from each class. New observations are classified using this line.

Classification/regression

K means

A clustering method that makes k clusters in which each observation belongs to the cluster that has its mean in the nearest locations from the observation.

Clustering

Hierarchical clustering

A type of cluster analysis that builds a dendrogram with a hierarchy of clusters. Pairs of clusters are merged to form clusters as they move up the hierarchy (agglomerative approach).

Clustering

Principal component analysis

An algorithm that converts high dimensional data into lower dimensional data while keeping important information as much as possible by orthogonal transformation.

Dimensionality reduction

270  The Smart Cyber Ecosystem for Sustainable Development Additional Input features ECG

Demographics

Clinical

Genomics

Normalization Imputation Reduction

Machine Learning

Cardiac Images

Whole input features dataset

Cardiac Diagnosis

Reduced input features dataset

Figure 14.4  Pipeline for building image-based machine learning models.

References 1. Thomas, H., Diamond, J., Vieco, A., Chaudhuri, S., Shinnar, E., Cromer, S., Perel, P., Mensah, G.A., Narula, J., Johnson, C.O., Roth, G.A., Moran, A.E., Glob Heart, Global Atlas of Cardiovascular Disease 2000–2016: The Path to Prevention and Control. 13, 3, 143–163, 2018. 4. 2. Global, regional, and national age-sex specific mortality for 264 causes of death, 1980–2016: a systematic analysis for the Global Burden of Disease Study 2016.GBD 2016 Causes of Death Collaborators. Lancet, 390, 10100, 1151–1210, 2017. 3. Gersh, B.J., Sliwa, K., Mayosi, B.M., Yusuf, S., Novel therapeutic concepts: the epidemic of cardiovascular disease in the developing world: global implications. Eur. Heart J., 31, 6, 642–8, 2010. 4. Gulshan, V., Peng, L., Coram, M., Stumpe, M.C., Wu, D., Narayanaswamy, A., Venugopalan, S., Widner, K., Madams, T., Cuadros, J., Kim, R., Raman, R., Nelson, P.C., Mega, J.L., Webster, D.R., Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs. JAMA, 316, 22, 2402–2410, 2016. 5. Krittanawong, C., Zhang, H., Wang, Z. et al., Artificial intelligence in precision cardiovascular medicine. J. Am. Coll. Cardiol., 69, 2657–2664, 2017. 6. Johnson, K.W., Torres Soto, J., Glicksberg, B.S. et al., Artificial intelligence in cardiology. J. Am. Coll. Cardiol., 71, 2668–2679, 2018. 7. Dawes, T.J.W., de Marvao, A., Shi, W. et al., Machine learning of three-dimensional right ventricular motion enables outcome prediction in pulmonary hypertension: a cardiac MR imaging study. Radiology, 283, 381–390, 2017. 8. Motwani, M., Dey, D., Berman, D.S., Machine learning for prediction of all-cause mortality in patients with suspected coronary artery disease: a 5-year multicentre prospective registry analysis. Eur. Heart J., 38, 500–507, 2017.

Impact of AI in CVD  271 9. Madani, A., Arnaout, R., Mofrad, M., Fast and accurate view classification of echocardiograms using deep learning. NPJ Digit. Med., 1, 1–8, 2018. 10. Poplin, R., Varadarajan, A.V., Blumer, K., Predicting cardiovascular risk factors from retinal fundus photographs using deep learning. Nat. Biomed. Eng., 2, 158–164, 2018. 11. Arterys cardio DL cloud MRI analytics software receives FDA clearance. In: editor book arterys cardio DL cloud MRI analytics software receives FDA clearance, 2017. Diagnostic and Interventional Cardiology. 2017. 12. Russell, S. and Bohannon, J., Artificial intelligence. Fears of an AI pioneer. Science, 349, 252– 252, 2017. 13. Verghese, A., Shah, N.H., Harrington, R.A., What this computer needs is a physician: humanism and artificial intelligence. JAMA, 319, 19–20, 2017. 14. Chen, J.H. and Asch, S.M., Machine learning and prediction in medicine-beyond the peak of inflated expectations. N. Engl. J. Med., 376, 2507–2509, 2017. 15. Yan, J., Wang, Z., Xu, L.J., Effects of new regional cooperative rescue model on patients with ST-elevation myocardial infarction. Int. J. Cardiol., 177, 494–496, 2017. 16. Reinke, L., Özbay, Y., Dieperink, W., The effect of chronotype on sleepiness, fatigue, and psychomotor vigilance of ICU nurses during the night shift. Intensive Care Med., 41, 657–666, 2017. 17. Maltese, F., Adda, M., Bablon, A., Night shift decreases cognitive performance of ICU physicians. Intensive Care Med., 42, 393–400, 2016. 18. Hu, J., Cui, X., Gong, Y., Portable microfluidic and smartphone-based devices for monitoring of cardiovascular diseases at the point of care. Biotechnol. Adv., 34, 305–320, 2016. 19. Wang, G., Hao, J., Ma, J., Jiang, H., A comparative assessment of ensemble learning for credit scoring. Expert. Syst. Appl., 38, 223–230, 2011.

15 Healthcare Transformation With Clinical Big Data Predictive Analytics Muhammad Suleman Memon1*, Pardeep Kumar2, Azeem Ayaz Mirani3, Mumtaz Qabulio4, Sumera Naz Pathan5 and Asia Khatoon Soomro5 Department of Information Technology, Dadu Campus, University of Sindh, Jamshoro, Sindh, Pakistan 2 Department of Computer Systems Engineering, Quid-e-Awam, University of Engineering Science & Technology, Nawabshah, Sindh, Pakistan 3 Department of Information Technology, Shaheed Benazir Bhutto Shaheed Benazirabad, Sindh, Pakistan 4 Department of Software Engineering, Faculty of Engineering, University of Sindh, Jamshoro, Sindh, Pakistan 5 Institute of Mathematics & Computer Science, University of Sindh, Jamshoro, Sindh, Pakistan 1

Abstract

Information technology is advancing day by day and it is used in almost every domain to transform it into a digital form. Big data management is a recent advancement to facilitate the management of a massive volume of structured, semi-structured, and unstructured data. The healthcare sectors are also producing a large volume of clinical data (e.g., medical reports, patient history files, and prescriptions) daily. This massive repository of data can be used to develop a predictive model to assist the healthcare experts to diagnose the diseases and suggest the proper medication to the patients. Though there are some serious concerns—legal and ethical—related to access healthcare data. This chapter focuses on providing a solution to find the hidden patterns from the clinical data by using state-of-art methods such as big data analytics, machine learning, and deep learning. The chapter will also discuss the various constraints for adopting predictive analytics in healthcare sectors. Keywords:  Big data, healthcare, clinical data, data analytics, predictive model, machine learning

15.1 Introduction Big data scope focuses on some important features called five Vs. The five Vs includes volumes (scales the data size), velocity includes the collection and analysis of the data which fixes the time limit, variety scales types of the data gathered from different resources, i.e., structured data, unstructured data, and semi-structured data, veracity involves the degree of validity of the data, and value involves the degree of data importance [1]. The world is transforming into the new idea of digitalization in every field. Many digital devices have been introduced through *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (273–286) © 2021 Scrivener Publishing LLC

273

274  The Smart Cyber Ecosystem for Sustainable Development which data is generated on a large scale. The data is being generated on the internet through social media, mobile phones, wearable devices, and Internet of Things (IoT). The data generated from multiple sources is large enough which requires huge computation power. The massive data generated is stored, processed, and analyzed for business applications. This forms a new trend called big data. Big data has been used in various fields such as transportation, for monitoring smart grid, improving business, and healthcare. Big data is used in many areas but in healthcare, the data is generated on a large scale. The data in healthcare is generated in the form of structured, semi-structured, and unstructured. According to the Institute for Health Technology Transformation [2, 3], the data in healthcare in the USA is generated in zettabyte and soon will reach yottabyte due to digitalization. Big data in the healthcare sector plays an important role due to its consistency and utility. The development of health recommender systems, clinical decision support systems as well as models for disease prediction help medical staff. The healthcare system generates a huge volume of data daily which is categorized as clinical data, patient and sentiment data, and pharmaceutical research and development data [4]. In China, guidance on introducing and standardizing the application and development of big data in healthcare started in 2016. It can be observed from this that China has implemented the concept of big data in its healthcare system [5]. Big data in healthcare has potential benefits to society. With the introduction of new technology such as the IoT sensors, social media, a global positioning system (GPS), and smartwatches, the people are considered as walking data generators [6]. Wearable sensors [7–11] and informal communication stages assume a key job in giving another technique to gather persistent information for productive medicinal services checking. Notwithstanding persistent patient checking, utilizing wearable sensors creates a lot of human services information. Furthermore, the client created human services information on interpersonal interaction locales that come in enormous volumes and are unstructured. The current medicinal services observing frameworks are not productive at removing significant data from sensors and person to person communication information, and they experience issues breaking down it viably. Diabetes and unusual circulatory strain (BP) are the two most normal and amazingly hazardous ailments that influence the functionality of the human body and increment the danger of cardiovascular maladies. Early identification and order of diabetes and BP into their classification is along these lines basic to treat the patient effectively. Figure 15.1 shows basic components of Big Data that include volume, velocity, variety and veracity. VELOCITY

VOLUME

Speed

Large Datasets

BIG DATA 4Vs

VARIETY Structured Semi Structured Unstructured

Figure 15.1  Big data 4Vs.

VERACITY Quality

Clinical Big Data Predictive Analytics  275

Data Sources

Data Storage

Batch Processing

Machine Learning

Real-Time Message Ingestion

Analytical Data source

Analytics and Reporting

Stream Processing

Figure 15.2  Big data architecture.

The Figure 15.2 depicts the working mechanism of Big Data. The massive volume of data is collected from various sources. After that data is stored. The data is processed in batch form over a period of time for further analysis.

15.1.1 Big Data in Health Sector The health system without the use of computer technology is very difficult when it comes to finding the record of the patient and the history of the patient. Big data in healthcare provides a great facility to gather data from various sources and produce meaningful insight from the data. In healthcare, data is generated in massive volumes daily. It is normally observed that the data is in a hard format which is quite difficult to process and decide. Doctors write prescriptions and there is no electronic record of those prescriptions. On the other hand, the digitized system helps to produce data which can be processed further. Big data is very helpful for timely delivery, accurate analysis, and cost-effective solution. Big data analytics [8, 12–14] helps healthcare professionals make quick decisions based on analyzed data, the cardiac patient once goes for bypass surgery or angioplasty is carefully monitored. The doctor can see the patient’s health status based on different factors, which helps the doctor to decide the treatment. The primary role of big data in healthcare is to gather huge volumes of data and analyze the complex data, which are difficult to manage using traditional hardware, or normal software, and other common management tools.

15.1.2 Data Structure Produced in Health Sectors The health sector produces data in various formats such as text, numbers, images, audio, and video. Due to the nature of the data produced by the health sector, it requires huge processing and intelligence to get meaningful information about the data. The health sector produces reports in textual format. The X-ray images, CT-scan images, MRI-images, ultrasound videos, ECG, and other data are produced by the health sector. The heterogeneous nature of big data in the health sector makes it difficult to process a large volume of the data. The data falls into two categories, i.e., structured and unstructured. The data formats text, images, SMS, email, audio, and video falls in the category of unstructured data. The other form of data is structured, which is stored in a relational database. Big data plays an important role to analyze data from various sources. The structure of the data in healthcare makes it difficult to analyze

276  The Smart Cyber Ecosystem for Sustainable Development data because the data is produced in different formats. The healthcare systems produce heterogeneous data, incompatible data, fragmented data, raw, and unstructured data [15].

15.2 Big Data Challenges in Healthcare There are tremendous challenges when we talk about big data in healthcare. In routine, the data is generated in a massive volume. The healthcare system produces multiple types of data daily such as textual data, images, audio, and video. Because of the different nature of the data, it requires intelligence to process all formats of the data. There are multiple challenges when we talk about big data in healthcare such as data structure issues, privacy and security issues, issues with standardization of data, lack of expertise, storage issues, real-time analytics issues, and managerial issues [15]. Healthcare systems can produce data of patients at various stages such as keeping records of a patient from birth to death. If the record is not computerized or is maintained by different hospitals on their own, it becomes difficult once the patient visits a different hospital which requires a proper history of the patient because the distributed data management systems separately maintained by hospitals do not help to conclude the data. It required a system to collect data from various sources and maintain a record of a patient at a single location. Big data management in the healthcare system will help government, hospital administration, doctors, patients, and pharmaceutical companies.

15.2.1 Big Data in Computational Healthcare Big data creates many opportunities to create a different application in the healthcare system. Many new approaches help normal users to check and provide quick assessment over the internet. In this context, many search engines are used to control and manage patient health status and give fast feedback over the internet. A few famous sites over the internet are American Well, MELIA, and many more which are considered an important source of online medical diagnose [16] the healthcare information system based on big data has been growing rapidly and is being adapted to medical information to derive important health trends and support timely preventive care. This research aims to evaluate organization-driven barriers in implementing a healthcare information system based on big data. It adopts the analytic network process approach to determine the aspect weight and applies VlseKriterijumska Optimizacija I Kzompromisno Resenje (VIKOR). The computational methods satisfy the huge amount of the data which was not easy ever before by convocational methods.

15.2.2 Big Data Predictive Analytics in Healthcare The COVID-19 pandemic has spread all over the world. On daily basis huge numbers of patients are admitted to hospitals. The data of the patients is manually maintained. Figure 15.3 shows the conventional methods and Artificial Intelligence approaches to smartly manage COVID patients. Nowadays, people are using social media a lot. There are various groups related to health where people share their problems with the people who are already suffering or getting any treatment. People with disease share their symptoms and the medication which they

Clinical Big Data Predictive Analytics  277 Conventional Methods

Data Undestanding Data Preparation Big Data

Human Intervention

Machine Learning - Regression Supervised Learning - Classification - Clustering Unsupervised Learning - Feature learning - Decision Making Reinforcement - Execution Learning Deep Learning Supervised Learning Unsupervised Learning

Experts

Semi-supervised Learning

COVID-19 Challenges

- Medical Image Analysis - Bioinformatics - Image restoration - Drug discovery - Toxicology - Natural language processing

Evaluation

Results Artificial Intelligence Approaches

Experinmental Treatment

Medical Equipment and Medical Imaging System Body Sensor Network

Previous Results

Geographical and Epidermatological Information

Figure 15.3  AI-based method to conquer COVID-19 [18].

are receiving. Social network has become a source of data generation related to healthcare. Besides, other sources help researchers to analyze data of the patients. Big data in the modern age is growing interest in different field’s life. Big data with emerging files can grow and create different applications of life. Machine learning is one of the effective approaches with emerging technologies of big data that created many new challenges and opportunities. Emerging fields such as IoT, image processing, machine learning, and other computer applications brought so many advancements in the healthcare industry. The new and latest technology related to the hardware as well as software tools has been playing a greater part today in hospitals to diagnose patient health. In big data concept, information gathering from different devices can further be investigated by applying machine learning methods and techniques for predicting the patient health status. This is a very impressive approach in big data science to improve the regular activities of patients and collect data over the cloud to apply different prediction methods. Machine learning [17] is applied to text, images, speech, and videos to analyze the data and get meaningful information from the data. In healthcare, the data comes from various sources such as examinations, laboratory tests, and medical imaging. By applying machine learning algorithms, data can be analyzed and can help doctors to predict future outcomes. Machine learning techniques can also help to make a recommender system that can guide doctors in recommending alternate therapy.

15.2.3 Big Data for Adapted Healthcare The big data concept has been adopted in the healthcare system due to the proximity and calamity of the user data. Big data involve in medical image analysis provides very high resolution and high throughput technology [20]. In data that involves images, it provides several statistical or machine learning methods that can be applied to evaluate image features for efficient analysis of the patient diagnosis to obtained good predictors. If the data

278  The Smart Cyber Ecosystem for Sustainable Development

Figure 15.4  MRI mages [19].

set contains a patient related to the observable subset of the data, then someone can compute the accuracy of the machine learning predictors. These predictors are further helpful to investigate the accuracy related to the different parameters. This provides an effective tool for medical image data analysis and diagnosis accuracy. Data through medical imaging is produced in large volumes from multiple sources such as MRI, CT scans, x-ray images and so on. Figure 15.4 shows the MRI images of tumors. The massive volume of medical imaging data can be stored through Big Data.

15.3 Cloud Computing and Big Data in Healthcare Cloud computing with big data institutes effective participation to improve the application of healthcare systems. It enhances the user experience and research community to arise effective usage for the patient diagnosis [21]. Big data applications services are migrating into cloud computing applications to perform some long-term solutions. In this context, it is very important to have safe and secure access to the data over the internet which is part of cloud computing applications. Big data storage is an important aspect of data collection that performs an effective role to apply statistical methods for patient records and investigates the healthcare data for future use. The data sets which are collected electronically for different IoT devices in big data science need to use cloud services for data processing, data collection, data storage, and other important aspects of the data representation. The flexibility with data representation will help to provide big data operations effective and rapid response. Table 15.1 illustrates the summary of commonly used Big Data Tools. The table shows the name of a tool, description, features, URL and price.

15.4 Big Data Healthcare and IoT IoT is things oriented, an internet-oriented, and semantic oriented network of objects [22]. These three paradigms (e.g., objects, internet, and semantics) are important to consider IoTenabled devices that can sense, process medical data for patient diagnosis, and further, it can be communicated in a real-time environment. Objects with additional internet connectivity having different functionalities can communicate with each other without any interference. It consists of a series of connected devices that scan share sensed data to optimize the performance of patient data. Make this sense for IoT devices to see, hear, think, and react to share data and information for making decisions based on obtained data. IoT devices smarter by underlying technologies include communication technologies, ubiquitous and pervasive computing, embedded technology, internet protocols, physical architecture, and many other

Clinical Big Data Predictive Analytics  279 Table 15.1  Big data tools. S. no.

Tool

Description

Features

URL

Price

1.

Hadoop

Apache Hadoop is one of the best big data frameworks. It facilitates distributed processing.

It uses an HTTP server for authentication. It provides flexibility in data processing and provides faster results.

https://hadoop. apache.org/ releases.html

Free

2.

HPCC

The HPCC uses a single platform, single architecture. It also uses a single programming language.

It uses less code to perform big tasks. It also provides optimization in parallel processing. It uses C++ programming language and can be extended using C++ libraries.

https:// hpccsystems. com/try-now

Free

3.

Storm

The Storm tool offers real-time computation.

It has some features like parallel calculations that can run across multiple machines. The tool is easy to use.

http://storm. apache.org/ downloads. html

Free

4.

Qubole

It is open source. It provides selfmanagement and selfoptimization. It allows the team to focus on the business outcome.

Uses a single platform. Provides security, compliance, and governance.

https://www. qubole.com/

Free

(Continued)

280  The Smart Cyber Ecosystem for Sustainable Development Table 15.1  Big data tools. (Continued) S. no.

Tool

Description

Features

URL

Price

5.

Cassandra

Used with massive data.

Provides a mechanism to replicate across multiple data centers. Provides a good mechanism for fault-tolerance. The replication mechanism offered in this tool provides an advantage of not losing data.

http://cassandra. apache.org/ download/

Free

6.

Statwing

It is easy to use a statistical tool.

Explores data in seconds. It provides good help to clean the data, and generate charts.

https://www. statwing.com/

Not Free

7.

CouchDB

It stores data in JSON that helps to process data over the web.

It is a single-node database. It uses Ubiquitous HTTP protocol and JSON data format. It helps in creating data replication across multiple servers.

http://comchdb. apache.org/

Free

8.

Pentaho

A good tool for providing insights into data.

Provides effective data visualization. Provides a mechanism to generate reports. It supports a good number of different big data sources.

http://www. pentaho.com/ download

Not Free

(Continued)

Clinical Big Data Predictive Analytics  281 Table 15.1  Big data tools. (Continued) S. no.

Tool

Description

Features

URL

Price

9.

Flink

Open source big data tool for stream processing.

It provides accurate results. It uses the stateful protocol and can recover from failure. This tool supports stream processing.

https://flink. apache.org/

Free

10.

Cloudera

It is the fastest, secure, easiest, and scalable platform.

It provides the functionality of multi-cloud. It can be easily deployed and managed over AWS, Microsoft Azure, and Google Cloud. It helps to train data models.

https://www. cloudera.com/

Not Free

11.

OpenRefine

Powerful big data analytics tool. It works with messy data.

It helps explore large datasets with ease. It also helps in generating data in various formats.

http://openrefine. org/download. html

Free

12.

RapidMiner

Open source platform.

It is used for data preparation and machine learning. It provides predictive data analytics. The data can be analyzed remotely.

https:// my.rapidminer. com/nexus/ account/index. html

Free

(Continued)

282  The Smart Cyber Ecosystem for Sustainable Development Table 15.1  Big data tools. (Continued) S. no.

Tool

Description

Features

URL

Price

13.

Hive

Open source big data software.

It provides SQL-like functionality to query data and model data. It compiles language with a map and reduces. It works only with structured data. It offers a JDBC interface

https://hive. apache.org/ downloads. html

Free

applications. Smart objects work on their relevant tasks with domain-specific information as well as an application that helps medical healthcare stockholders to diagnose efficiently [23]. IoT brings new trends to improve living standards and contribute to providing a rapid response to real-life medical applications. Business and economy can grow with the ratio of daily life interest by the implementation of smart sensor technology. For instance, home automation brought more luxury to get repaid updates of patient health. Big data creates opportunities by obtaining real-time data for IoT devices which helps to attend long-term data analysis for patient healthcare. The potential growth of smart physical devices arises market demand and potential health monitoring needs of the patient. However smart devices need to develop and provide patient statistical data that focus on a few important parameters such as availability, reliability, and security among the heterogeneity of the devices [24]. The efficiency of physical devices will be led to satisfying patient health. Moreover, architecture, protocols, and the internet are important involvement in IoT platform support more suitable for customer and business satisfaction. Besides, the internet will help to provide data over the cloud for long term accessibility of data for patient diagnosis. IoT devices communication over the cloud is dense and remote areas where accessibility of the medical representative is not easy [25]. So many different devices are connected over the internet having different properties, architecture, and protocols so it is necessary to build suitable for reliable and efficient communication.

15.5 Wearable Devices for Patient Health Monitoring When we talk about the health of a patient, the two most important things are taken into primary consideration—sugar and blood pressure. These two are very dangerous [7] and very important to take into consideration for further diagnosis. The healthcare professionals try to monitor the blood pressure and sugar periodically. To diagnose the patient efficiently, it is important to monitor the blood pressure and sugar regularly. With the advancement of technology, the industries are relying more on automated machines to do the job rather than doing things manually. The health sector is also using technology devices to diagnose different parameters of patient health. Wearable devices play an important role to do this job. Wearable

Clinical Big Data Predictive Analytics  283 1 Breathing Rate Sensor: Detects chest expansoion as a result of breathing 1

Breathing Rate Sensor 2 Heart Rate Sensor

3 Calf Muscle Sensors 4

2

Heat Rate Sensor: Detects pressure changes due to expansion/contraction of the wrist due to arterial blood flow

3

Calf Muscle Sensor: Detects contraction/expansion of the muscle resulting when the calf undergoes activity

4

Gait Sensor: Detects the ground reaction force that is induced when walking

Ground Reaction Force Sensors

Figure 15.5  Summary of wearable sensors [30].

devices [4, 26–29] are used in patients with critical conditions or elderly patients. These devices are deployed in the human body such as skin or employed over the human body such as on wrist and other body parts to collect data to carefully monitor various health-related issues in elderly and critical patients. Wearable devices help to monitor patient blood pressure, heart rate, oxygen, etc. In order to monitor critical patients various sensors are attached to patients body to collect the data such as breathing rate, heart rate, blood pressure etc. Figure 15.5 shows various wearable sensors are attached to human body to collect important data of the patient.

15.6 Big Data and Industry 4.0 The Industry 4.0 term is used in Germany where they planned a new strategic policy for the economy of the country. The high-tech industry is based on internet-connected devices having a huge digital market [31]. Industry 4.0 is considered more advanced, fast, reliable, and secure for the improvement of the global economy. This industry is called the fourth revolution which is based on the world’s most advanced technologies for the reliable connectivity of global business. This industry is based on the cyber-physical system having certain methods of data automation and revolving the manufacturing technologies all over the world [32]. It includes IoT, cloud computing, cognitive computing, and cyber-physical systems. It is a smart industry in a digital sense which makes all possible cyber-physical activities. This helps to monitor and control the activities of the physical cyber system within the manufacturing area.

15.7 Conclusion In this chapter, we have discussed the role of big data in healthcare transformation. We have discussed big data, its architecture, applications in healthcare, the IoT, Industry 4.0, and various tools used in big data.

284  The Smart Cyber Ecosystem for Sustainable Development

References 1. Rahul, K. and Kumar, R., ScienceDirect ScienceDirect Data Life Cycle Management in Big Data Analytics. Procedia Comput. Sci., 173, 2019, 364–371, 2020. 2. Lv, Z. and Qiao, L., Analysis of healthcare big data. Futur. Gener. Comput. Syst., 109, 103–110, 2020. 3. Bahri, S., Zoghlami, N., Abed, M., Tavares, J.M.R.S., BIG DATA for Healthcare: A Survey. IEEE Access, 7, 7397–7408, 2019. 4. Galetsi, P. and Katsaliaki, K., Big data analytics in health: an overview and bibliometric study of research activity. Health Info. Libr. J., 37, 1, 5–25, 2020. 5. Rizwan, A. et al., A Review on the Role of Nano-Communication in Future Healthcare Systems: A Big Data Analytics Perspective. IEEE Access, 6, 41903–41920, 2018. 6. Hadi, M.S., Lawey, A.Q., El-Gorashi, T.E.H., Elmirghani, J.M.H., Patient-Centric Cellular Networks Optimization Using Big Data Analytics. IEEE Access, 7, 49279–49296, 2019. 7. Ali, F., El-sappagh, S., Islam, S.M.R., Ali, A., An intelligent healthcare monitoring framework using wearable sensors and social networking data. Futur. Gener. Comput. Syst., 114, 23–43, 2021. 8. Sahoo, P.K., Mohapatra, S.K., Wu, S.L., Analyzing Healthcare Big Data with Prediction for Future Health Condition. IEEE Access, 4, 9786–9799, 2016. 9. Lin, K., Xia, F., Wang, W., Tian, D., Song, J., System Design for Big Data Application in EmotionAware Healthcare. IEEE Access, 4, 6901–6909, 2016. 10. Firouzi, F., Farahani, B., Ibrahim, M., Chakrabarty, K., Keynote paper: From EDA to IoT eHealth: Promises, challenges, and solutions. IEEE Trans. Comput. Des. Integr. Circuits Syst., 37, 12, 2965–2978, 2018. 11. Hughes, J. and Iida, F., Multi-functional soft strain sensors for wearable physiological monitoring. Sensors (Switzerland), 18, 11, 3822, 2018. 12. Moorthy, U. and Gandhi, U.D., A Survey of Big Data Analytics Using Machine Learning Algorithms. In HCI challenges and privacy preservation in big data security, pp. 95–123. IGI Global, 2017. 13. O’Connor, S. (2018). Big data and data science in healthcare: What nurses and midwives need to know. J. Clin. Nurs., 27, 15–16, 2921–2922, 2018. 14. Sethi, D. and Anand, J., Big data and WBAN: Prediction and analysis of the patient health condition in a remote area. Eng. Appl. Sci. Res., 46, 3, 248–255, 2019. 15. Kruse, C.S., Goswamy, R., Raval, Y., Marawi, S., Challenges and Opportunities of Big Data in Healthcare: A Systematic Review. JMIR Med. Informatics, 4, 4, e38, 2016. 16. Chen, P.T., Lin, C.L., Wu, W.N., Big data management in healthcare: Adoption challenges and implications. Int. J. Inf. Manage., 53, 102078, 2020. 17. Smiti, A., When machine learning meets medical world: Current status and future challenges. Comput. Sci. Rev., 37, 100280, 2020. 18. Jamshidi, M., Lalbakhsh, A., Talla, J., Peroutka, Z., Hadjilooei, F., Lalbakhsh, P., ... Mohyuddin, W., Artificial intelligence and COVID-19: Deep learning approaches for diagnosis and treatment. IEEE Access, 8, 109581–109595, 2020. 19. Zheng, H. et al., Multi-Contrast Brain MRI Image Super-Resolution With Gradient-Guided Edge Enhancement. IEEE Access, 6, 57856–57867, 2018. 20. Viceconti, M., Hunter, P., Hose, R., Big Data, Big Knowledge: Big Data for Personalized Healthcare. IEEE J. Biomed. Heal. Informatics, 19, 4, 1209–1215, 2015. 21. Zhang, X. et al., Proximity-aware local-recoding anonymization with MapReduce for scalable big data privacy preservation in cloud. IEEE Trans. Comput., 64, 8, 2293–2307, 2015.

Clinical Big Data Predictive Analytics  285 22. Forkan, A.R.M., Khalil, I., Ibaida, A., Tari, Z., BDCaM: Big Data for Context-Aware Monitoring—A Personalized Knowledge Discovery Framework for Assisted Healthcare. IEEE Trans. Cloud Comput., 5, 4, 628–641, 2015. 23. Wuest, T. et al., Machine learning in manufacturing : advantages, challenges, and applications. Prod. Manuf. Res., 3277, 1–23, 2016. 24. Mirani, A.A., Memon, M.S., Bhati, M.N., Soomro, M.A., Rahu, M.A., Taxonomy of ubiquitous computing: Applications and challenges. 2017 Int. Conf. Inf. Commun. Technol. ICICT 2017, vol. 2017-Decem, pp. 202–208, 2018. 25. Stolpe, M., The internet of things: Opportunities and challenges for distributed data analysis. Acm Sigkdd Explorations Newsletter, 18, 1, 15–34, 2016. 26. Pashazadeh, A. and Navimipour, N.J., Big data handling mechanisms in the healthcare applications: A comprehensive and systematic literature review. J. Biomed. Inform., 82, March, 47–62, 2018. 27. Galletta, A., Allam, S., Carnevale, L., Ali Bekri, M., El Ouahbi, R., Villari, M., An innovative methodology for Big Data visualization in oceanographic domain. ACM Int. Conf. Proceeding Ser., vol. 15, pp. 103–107, 2018. 28. Nambiar, R., Bhardwaj, R., Sethi, A., Vargheese, R., A look at challenges and opportunities of Big Data analytics in healthcare. Proc. - 2013 IEEE Int. Conf. Big Data, Big Data 2013, pp. 17–22, 2013. 29. Hossain, M.S. and Muhammad, G., Healthcare Big Data Voice Pathology Assessment Framework. IEEE Access, 4, 7806–7815, 2016. 30. Choudhury, T., Chhabra, A.S., Kumar, P., Sharma, S., A recent trends on Big Data analytics. Proc. 5th Int. Conf. Syst. Model. Adv. Res. Trends, SMART 2016, pp. 225–231, 2017. 31. Aceto, G., Persico, V., Pescapé, A., Industry 4.0 and Health: Internet of Things, Big Data, and Cloud Computing for Healthcare 4.0. J. Ind. Inf. Integr., 18, February, 100129, 2020. 32. Wan, J., Tang, S., Shu, Z., Li, D., Wang, S., Imran, M., Vasilakos, A.V., Software-defined industrial internet of things in the context of Industry 4.0. IEEE Sens. J., 16, 20, 7373–7380, 2016.

16 Computing Analysis of Yajna and Mantra Chanting as a Therapy: A Holistic Approach for All by Indian Continent Amidst Pandemic Threats Rohit Rastogi1*, Mamta Saxena2, D.K. Chaturvedi3, Mayank Gupta4, Mukund Rastogi5, Prajwal Srivatava5, Mohit Jain5, Pradeep Kumar5, Ujjawal Sharma5, Rohan Choudhary5 and Neha Gupta5 Physics and CS Department, DEI Agra and Associate Professor, Department of CSE, ABESEC, Ghaziabad, U.P., India 2 DG, Ministry of Statistics, Government of India, Delhi, India 3 Electrical Engineering, Dayalbagh Educational Institute, Agra, U.P., India 4 Tata Consultancy Services, Japan 5 Department of CSE and CEIT, ABESEC, Ghaziabad, U.P., India

1

Abstract

The 21st century has brought many unpredicted changes in our lives. Science in 18th, 19th, and 20th century was developed in different dimensions and spread rapidly to make human life more comfortable. Unknowingly, its vision was blind so it became uncontrolled. It bypassed the basic human emotions, ethics, and responsibilities, and consequently, the happiness, prosperity, and peace of mind from our lives were lost. The ancient Hindu Vedic life style was completely scientific and they experienced and searched the formula of being contented and making themselves optimistic and healthy in all adverse circumstances, less resources and less. The present article is a trial to establish the scientific way of living in Hindu religion. The experimental analysis and logical explanation have been presented to readers to understand this fundamental aspect in view of Covid-19 and its various effects on physical, mental, social, and spiritual health of individual, family, and society. The present paper is a hope for humanity amidst pandemic threats surrounded over globe. Keywords:  Yajna, Mantra, Covid-19, Om, happiness, data analytics, computational intelligence

16.1 Introduction 16.1.1 The Stats of Different Diseases, Comparative Observation on Symptoms, and Mortality Rate Disease: Rabies Symptom—animal-like behavior, chances of death is 100%; Ebola Symptoms— weakness and fever, chances of death is 90%. *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (287–306) © 2021 Scrivener Publishing LLC

287

288  The Smart Cyber Ecosystem for Sustainable Development Virus: Marburg Symptoms—problem in intestines, death within 10 days, chances of death is 88%; Nipah Symptoms—death after mental confusion, chances of death is 75%; Crimean Congo Identification—blue nose and mouth and their bleeding, chances of death is 40%; SARS Identification—heavy breathing, chances of death is 36%; Zika Identification—joint pain and skin rash, chances of death is 20%; Influenza Identification—throat irritation and pain, chances of death is 13%; Corona Identification—respiratory tract infection, chances of death is 2% [3, 32, 44]. Do you know that in these last two and a half months, more than 75 thousand people were infected in China due to coronavirus out of which only 2,200 people died? Do you also know that in these last 4 months in the US, more than 22 million people were infected with influenza and more than 16 thousand people died? So how did coronavirus become scarier than influenza? This is called the power of media and trust in it. In addition, it is called to incarcerate the people of the whole world in the media. Media war is a soft war of today. Virus-related cases and countries affected by them are as HIV—Congo, Nipah—Malaysia, Ebola—Sudan, Bird flu—Hong Kong, and Dengue—Manila, and Corona—China [22].

16.1.2 Precautionary Guidelines Followed in Indian Continent Just because schools are closed, it does not mean people avoid getting that compulsive travel and holiday bug. Holidays will come next year too why try your luck with Corona especially with children. Marriage functions, birthday parties, etc., can wait. Do not try your luck and that bravado that nothing will happen to me. The next 30 days will be most crucial in medical history of India. Take all precautions while at home and while outside for any important work. Precaution is not to panic. Be a responsible citizen by following and educating others to remain careful for next 1 month. Take this seriously and avoid all unnecessary social contacts [2, 20] (as per Figure 16.1).

Figure 16.1  The precautionary steps to be followed by individuals in quarantine time.

Yajna and Mantra Chanting as a Therapy  289

16.1.3 Spiritual Guidelines in Indian Society 16.1.3.1 Spiritual Defense Against Global Corona by Swami Bhoomananda Tirtha of Trichura, Kerala, India Ayurveda doctors are saying that we can save ourselves from coronavirus with the following guidelines. 1. Boil black peppers in water and add lemon juice, drink as soon as you come home. It kills the virus. 2. Drink warm water with cinnamon and basil leaves daily. No normal water and cold water. 3. Bath with salt water [3, 7, 8].  4. You can use eucalyptus oil as hand sanitizer. You can also inhale small quantities everyday. It kills the virus [51, 52]. 5. Take more lemons with hot water and turmeric [64, 65]. 

16.1.4 Veda Vigyaan: Ancient Vedic Knowledge Veda Vigyaan: In the Yajna Havishyanna (powder form of various herbs), the Static Sun energy is micronized and multiplied with the sound energy of Mantra and gives it to the Yagya Agni, which becomes a conductor and makes the energy a cosmopolitan and as a conductor; it is spread in the environment and destroys the germs and viruses. The century life of Vedika was found by its regular use. All researchers should think to use this unfailing weapon for corona. Veda Vigyaan: Vedabhagavan (The God Veda) says that the sun rays, especially in the morning and evening, are germicidal. We should use this effective weapon for corona [19, 21].

16.1.5 Yagyopathy Researches, Say, Smoke of Yagya is Boon Usually, in science, smoke is considered to be poisonous and bad for health but the Vedic aspect of science says that the smoke generated by Yajna activity is good for human and environmental health. It is useful for plants, kills bacteria and pathogen, and purifies the inner respiratory organs of human body like lungs. It is much effective in different diseases and the smoke of herbs generated by combustion of medicinal plants is acting therapeutically [4, 9]. The subject suffering from different ailments is asked to lie or sit down near the Yajna process and inhale deeply the smoke of Yajna. Many diseases are being cured nowadays by this therapy, popular termed as Yagyopathy [10–12]. The need of today is the amalgamation of Spirituality and Science. Dr. Pranav Pandya, the chancellor of DSVV Haridwar and current head of all world Gayatri Parivaar, Shantikunj Hardwar, has been pioneer of this field and wish that science should come forward and take responsibility to prove the validity of this ancient Vedic rituals [13, 18]. Indian culture is filled with scientific practices. It is said that the smoke from the YagyaHawan is a boon for health and can also cure depression, disorders, and many more diseases. Yagya-Hawan uses a variety of medicinal plants [53].

290  The Smart Cyber Ecosystem for Sustainable Development Dr. Vinay Pathak, VC of AKTU Lucknow, has also appreciated by DSVV Haridwar and volunteers of Yagyopathy Research to bring the ancient glory of this Indian Science [54, 63].

16.1.6 The Yagya Samagri To perform Havan to treat diseases like dengue or viral, the ingredients needed are as follows: Swertia, Green chiretta, Sweet wormwood, Kapoor tulsi, Purple Tephrosia, Devil’s tree leaves, Cultivated Licorice, Tinospora Cordifolia, Anantmool, Cannabis Sativa, Picrorhizakurroa, Millettiapinnata seeds, Pointed gourd leaves, Neem, and all others [18, 20]. All the possible “common Havan ingredients-1” should be mixed together to perform the Havan, and the above 14 ingredients should be mixed well together and powered. The powder should be mixed with lukewarm water and should be taken by the patient, one spoonful every morning and evening [14].

16.2 Literature Survey 16.2.1 Technical Aspects of Yajna and Mantra Therapy Mantra is a unique sound which has been found by different sages and Rishis of Vedic times in the supernatural stage of meditation while listening to it in subtle domain in cosmos, and then, they tried to arrange the syllables in form a Mantra [23, 24]. If anyone individually chants mantra, then oneself only gets benefit of it, but for mass access and benefit of this mantra energy in large surface, the sound and light energy are associated with it in ritual of Yajna. This process makes the mantra energy spread at larger area and more powerful [16, 17].

16.2.2 Mantra Chanting and Its Science During Yajna process, how the chanted Mantra sound energy vibrations reach to global domain and cosmos to spread all around, this whole concept can be understood by radio broadcasting process in present fundamental physics [25]. The frequency of sound waves is less and wavelength is more so due to lack of much energy; these waves are not able to transmit for farther range. Also, the sound waves need some medium for transmission [45, 54]. In current science, these sound waves are super imposed on carrier waves over radio waves. Frequency of radio waves is similar to frequency of EM radiations. Being equipped with high frequency waves, radio waves can move over cosmos so consequently sound waves can also travel to far area after super imposing these on these radio waves. This is called simple modulation methods, and by this method, the present radio broadcasting station works [62].

16.2.3 Yagya Medicine (Yagyopathy) Yugrishi and founder of Shantikunj, Sh. Sriram Sharma Acharya Ji, stated in his literature (Yajna, Ek Samgra Upchaar Prakriya, Vangmay-26, Section 1.7 and 1.8) that the enlightened

Yajna and Mantra Chanting as a Therapy  291 fire during Yajna generates heat and light energy of this radiation or wave is electromagnetic by nature. These waves act as carrier waves for sound waves. The sound waves generated at time of Mantra chanting are amplified and transmitted on the electromagnetic waves after superimposing on them [5, 6, 61]. This is the underlying secret that the Mantra is made more powerful and might, along with ready to release effects while adding it with heat and light energy during Agnihotra or Homa Therapy [1, 15].

16.2.4 The Medicinal HavanSamagri Components 16.2.4.1 Special Havan Ingredients to Fight Against Infectious Diseases To perform the ritual against environmental pollution, the contents required are as follows (the biological names of some plants have been mentioned): Agarose, Anantmul (a long unending root), Apamarg, dry leaves of Mango tree, Amla (Phyllanthusemblica), Big cardamom, Indrajo (a deciduous shrub), Big Kateri, Camphor, Raisin, Kooth, Saffron, Giloy (Tinospora Cordifolia), Commiphorawightii, Supfur, Gandtrin, Jasmine, Amaranth root, Sandalwood powder, dried date, Jatamansi, Nutmeg, Barley, Crape jasmine, Taj, Himalayan silver fir, Tulasi, Indian bay leaf, Davna, Cinnamon, Cedar powder, Cynodondactylon, five leaved chaste leaf, Cyperusrotundus, Indian Jalap, dried leaves of Neem, Nagkesar, Nerium leaves, perfumed cherry, stem bark, Calamus, Bala (SidaCordifolia), Bakuchi (Psoraleacorylifolia), Eucalyptus, Indian goldthread, Holy Basil, Maulshri (Mimusopselengi), Indian madder, Indian Copal, Styrax benzoin leaves, Lodhra (SymplocosRacemosa), Cloves, Cirrus bark, Mustard, Sal, Valerian leaves, Sarj, Betel, Turmeric, Haritaki (Terminaliachebula), Cow’s milk, Jaggery or sugar or honey, Barring jaggery, and melted butter. All the above-mentioned contents should be taken in equal measurements and powered together. Thereafter, all the proper ingredients must be mixed well and saved [26, 27]. During Havan, the “common Havan ingredients-1” and one-tenth of melted butter and one-tenth of jaggery or sugar or honey should be mixed, and the Gayatri Mantra should be chanted. The Havan should be performed either during sunrise or sunset. This is the “Sanshivela” time which is considered safe and effective for the Havan. All the above-mentioned ingredients should be taken in equal quantities. Most of the items are easily available which are also used in the Basic Havan [55, 58].

16.2.5 Scientific Benefits of Havan A research conducted by the National Botanical Research Institute has found that medicinal smoke generated during Pooja and Havan purifies the environment by destroying harmful bacteria and reducing the possibility of spreading the disease to a great extent. Burning wood and medicinal herbs, which are called Havan materials in common language, bring purity in the atmosphere and burn viruses and similar kinds of beings up to 94% [28, 57]. In order to confirm the research and to check the scientific effect of the smoke of the incense materials on the atmosphere, it was used in a closed room. In this experiment, a mixture of more than five herbs is used. This incense was sourced from Guru Kul Kangri Haridwar Institute. The room environment before and after the Havan was extensively

292  The Smart Cyber Ecosystem for Sustainable Development analyzed and tested and was found that the medicinal smoke generated from the incense reduced the amount of harmful bacteria present in the air by up to 94%. The effect of this medicinal smoke lasts for 30 days, and during this period, toxic germs do not remain alive. The action of smoke not only has a good effect on the health of man, but this experiment has also proved to be very effective in agricultural [29, 56]. Scientists say that in earlier experiments, it was found that the smoke of medicinal Havan can also get rid of harmful bacteria that damage the crop. Compared to all kinds of medicines given to humans, medicinal herbs and medicated Havan smoke are more beneficial to fight against many diseases and it does not cause any harm, while the medicines have some side effects. Smoke is directly effective in human body and this system is cheaper and more durable than medicines [59, 60].

16.3 Experimental Setup Protocols With Results The happiness index of students after the Yagya was found to be increased. Although, the students in the age of 20 had happiness index slightly more than others. The highest was found to be in a 21-year-old student. It shows how people of different age can have different happiness indexes. In all cases, it was seen that the happiness index was more as compared to before. This analysis carried out by IoT-based sensors. These sensors helped us to obtain accurate data (as per Figure 16.2). The happiness index of people from Fatehpur had also increased after the Yagya. It shows that the place does not matter at all. In almost all cases, the index was more as compared to before. It proves that the Yagya has the same effect irrespective of the place it is held in. The sensors used were IoT equipped which gave us precise data for more informative analysis (as per Figure 16.3). The highest happiness index value was found to be in the age of 20s. The least was in students in the age of 21. Although after the Yagya, the happiness index had decreased in

Happiness Index of ABESEC Hostel Students, Ghaziabad, NCR, India (Metropolitan region)

Happiness Index Value

90 80 70 60 50 40 30 20 10 0 Before and After analysis of students Before

Figure 16.2  The happiness index of metro city.

After

Yajna and Mantra Chanting as a Therapy  293 Happiness Index of people from Fatehpur, UP, India (Rural region)

Happiness Index Value

120 100 80 60 40 20 0 Before and After analysis of people Before

After

Figure 16.3  The happiness index of rural city.

most of the students as compared to before. But, it also hints us how students of today’s generation can be made happy. By the use of AI and ML, the analysis became very easy to be carried out. The world is now adopting these methods so that complex things can be made easier (as per Figure 16.4). In this case, the highest happiness index is of people in the age group 30–39, before Yagya. After Yagya, the highest happiness index was in the people of age group 30–39. The lowest was in people win the age group 70–79, before Yagya. The highest index after Yagya was in the age group of 50–59. A slight drop in the index was found to be in every case. AI and ML helped us to combine the data, which made the work very easier for us to compile. The data, being abstract and huge, could have cost us a lot of time. But in such cases, advanced technologies proved to be very beneficial (as per Figure 16.5).

Happiness Index vs age of hostel students of ABESEC, Ghaziabad, NCR, India (Metropolitan region)

Happiness Index Value

80 70 60 50

62.5

68.11 57.5

56.69

40

58.35 50.125

30 20 10 0

19

20

21 Age of students Before

After

Figure 16.4  The happiness index vs. age correlation of metro city.

294  The Smart Cyber Ecosystem for Sustainable Development Happiness Index vs age of people of Fatehpur, UP, India (Rural region)

Happiness Index Value

70 60 50

56 53.5

40

57.91

52.44

54 45.36

38.38

30

36.57

39.36 33.13

38.0836.21

50-59

60-69

32.9134.5

20 10 0

15-19

20-29

30-39 40-49 Age group of people Before

70-79

After

Figure 16.5  The happiness index vs. age correlation of rural city.

It shows the radiation coming out from various electronic devices. Some had high value, whereas some had a low value. Highest was found to be in the case of laptops and lowest was found to be in chargers, watches, and headphones. After performing the Yagya, it was seen that the radiation index had decreased a lot. Hence, Yagya can be a measure to fight against increase of radiation. This data was compiled with the help of big data analysis. Big data helped in analysing and extracting all the data systematically to get further information (as per Figure 16.6). The results were found to be alike of that in ABESEC Hostel. The radiation value had decreased a lot after the Yagya. It shows that Yagya was very helpful irrespective of the place it was performed in. These radiations, which are very harmful for us, can be tackled. As the data was too large and abstract, hence AI and ML helped to combine it and get the required result. The analysis was carried out through IoT systems (as per Figure 16.7).

Radiation Value

Radiation Analysis in ABESEC Hostel, Ghaziabad, NCR, India (Metropolitan region) 20 18 16 14 12 10 8 6 4 2 0

Mobile Phone Headphone

Laptop

Watch

Types of Electronic gadgets used Before

After

Figure 16.6  The radiation analysis of different gadgets in metro city.

Charger

Speaker

Yajna and Mantra Chanting as a Therapy  295 Radiation Analysis in Fatehpur, UP, India (Rural region)

Radiation Value

25 20 15 10 5 0 Before and After analysis of people Before

After

Figure 16.7  The radiation analysis of different gadgets in metro city.

Sample - Location Distribution

Location ABESEC, Ghaxiabad Gayatri Shaktipeeth HariharGanj Fatehpur ABESEC, Ghaxiabad 20

Total Subjects 99

Gayatri Shaktipeeth HariharGanj Fatehpur 79

Figure 16.8  All sample location distribution visualization.

16.3.1 Subject Sample Distribution For the study, there were a total of 99 subjects chosen at random from different backgrounds. For the analysis, normally, subjects belonging to middle class, who have self-interest in spiritual activities and are self-motivated, have been chosen. Subjects volunteered themselves for the study (as per Figure 16.8).

16.3.1.1 Area Wise Distribution For the study purpose, mainly experiments are conducted at two places: one city is industrial and has lifestyle of metro cities, which is Ghaziabad, and the other is normal city, Fatehpur.

16.3.2 Conclusion and Discussion Through Experimental Work The happiness index was calculated by happiness calculator gadget which has certain sensors, and the subjects were asked to put their point finger and thumb of the left and right

296  The Smart Cyber Ecosystem for Sustainable Development hands on this device. The readings were recorded pre- and post-Yajna ceremony, and graphs were visualized. It was found that, for both the hands, post-Yajna stress level recorded was less than the prequel of the activity. Please refer to Figure 16.9, where one can easily find that, in the right hand of the subjects, the happy cool and harmony states of subjects who are both male and female after the Yajna were more than before the Yajna. It means that the activity has impacted a positive and motivating impression on them. Same was the observation for left hand too. Similar, it can be remarkably noted that the subjects with low, medium and active stress level were drastically reduced after the Yajna process and Mantra chanting. The therapy has proven a clear cut vision for improvement on the mental fitness and subjects’ anxiety level, tension, depression, and fatigue levels. The summary of the different subjects with active, medium, and low stress along with happy and cool state of mind can be visualized by graph presented below. The above graph (please refer to Figure 16.10) demonstrates the data visualization of effect on the brain hemispheres after and previous to the Yajna activity. Both globes of the skull are being affected by this process, and it has been clearly depicted for the youths aged for 21 to 40 years; the effect of the Yajna activity was highest. The activity produced sharp rise in the kids and adolescents of less than 20 years but the medium variation was recorded for the both globes of skull in adults and very less effectiveness in old people. The following observations and reasons for aforesaid results can be understood. Since, the kids and school going chaps have less tension and pressure in their lives and they usually release their mental pressure while playing and soft upbringing by parents. The maximum pressure bearing candidates in our society are middle age youths of 21 to Happiness Index Female

Male

Happiness: Left - Before Yagya

30

19

30 4

20 10

10

7

0

5 Happy, Cool and Harmony State

27

19

8 Low Stress Level

Subject Count

40

Medium Stress Level

20

0

6 Happy, Cool and Harmony State

28

14 4

24

10 11

11

Happy, Cool and Harmony State

Low Stress Level

40

4 16

9 Low Stress Level

11

13 Acute Stress Level

Acute Stress Level

13

30 20

12

10 0

Medium Stress Level

Medium Stress Level

Happiness: Right - After Yagya

30

8

20

Acute Stress Level

18

10

11

0

Happiness: Right - Before Yagya

10

Subject Count

Subject Count

40

Happiness: Left - After Yagya

Subject Count

Gender

12

11

12

Happy, Cool and Harmony State

Low Stress Level

32

4 Medium Stress Level

Acute Stress Level

Figure 16.9  Left- and right-hand happiness index analysis before and after Yagya activity.

Yajna and Mantra Chanting as a Therapy  297 Avg Improvement Measure Names Avg. Improvement-Left Avg. Improvement-Right

12.250

12 10.917 10

Value

8 7.625 6

5.216 4.946

4

4.100 2.600

2 0 Less than 20 yrs

21-40 yrs

41-60 yrs

More than 60 yrs

Figure 16.10  Average improvement left- and right-hand analysis after Yagya activity.

40 years. This 21- to 40-year stage is the most crucial time of one’s life. This is the time when one has to clear the board exams, set his/her career, get a good admission in reputed institute, complete his/her studies, and get job. Also, the peer pressure and effects of parents and social impressions are also to be dealt with. In this time, the affairs, physical infatuations, attraction, and much kind of diversions like friends, social media, and drugs addiction create a lot of anxiety, depression frustration, and mental pressure. The Yajna and Mantra have revealed the best effect on this age group than any other citizen of the society because in such environment, these subjects gets the maximum relief and relaxation from their personal chaos and issues (as per Figure 16.10). The variation in the left and right hand reading is due to the reason that gland’s release of hormones from both the hands may be different and the posture of sitting, inhaling, and exhaling, and mental level also matters on the reading values. The stress is also governed by many parameters like subjects’ physical fitness, mental health, social relationship, upbringing, inner strength, and confidence. In meditation, people also travel to their past and some of them take worries for future. Now, what kind of story one weaves in one’s mind, the secretion occurs in similar ration from both the hands and it may vary some times as our thoughts jump very fast time and again and we deviate from one end to other in puzzles of these thoughts. But it is sure and clearly reflected through the graph that the Yajna and Mantra chanting is a very affirmative spiritual therapy where the guided and unguided meditation also helps and effectively brings the subjects back in cool state of mind.

16.4 Future Scope and Limitations There may be sufficient limitations in the study as it was conducted in a constrained environment, and mostly, the subjects do not follow the whole protocol instructed by the trainers. The sitting postures, breathing techniques, the thought process, and self-regulation practices are the key components which bring variation among the experiments [34, 35]. Environmental effects, family conditions, and physical or economical crisis are such factors along with marital relationship which affect one’s mental and social health a lot.

298  The Smart Cyber Ecosystem for Sustainable Development Consideration of all these parameters is important and the readings can be taken in more frequency for longer duration of time. The subject strength can be increased for more authenticated data and more refined sensor based gadgets can be used to get more accurate readings of the happiness and stress levels of the subjects under study [36, 37].

16.5 Novelty The manuscript establishes uniquely the power of Ayurveda in a novel way. There is a description of galaxies, powders, and pellets. All the medicines used in them are potent but also strong. Due to seniority, their digestion is not done properly, and as a result, they do not get the benefit that should be got. Agnihotra overcomes this difficulty. In that, there is no question of digestion. Delicately, the quality runs in the blood through the breath, affects the organisms, is unquestionable. It nurtures children like hair, aged, male, female, all equally (Yajna: a holistic healing process, Yugrishi Shriram Sharma Acharya) [38]. Medicines burnt during Yagya produce fumes (radiations). These are not only fumes of nutritious and healthy substance which have the same ability to make genes pure and health as the ability to contaminate toxic substance, but the sound of mantras waves the air of purity all over the sky. Therefore, the benefit of Yagya, more or less benefits the entire world the same way the radiations of bombs pushes the whole world toward pain and madness (Yagya: it is a holistic healing process, Yugrishi Shriram Sharma Acharya) [39, 40]. Yagya Shakti has the power for killing the germs and diseases. This same amount of simple, comprehensive and inexpensive method has not yet been discovered (Yagya: it is a holistic healing process, Yugrishi Shriram Sharma Acharya). Whose body spreads the good smell of perishable googal; that subject does not suffer from tuberculosis (lung disease). He does not get the curse of anyone either. All kinds of Yakshama disease (tuberculosis) are scared of him like a deer and run away with fear [30, 41].

16.6 Recommendations Need of removal of the poor attitude of West toward Hindu culture: When Hindus were wishing each other with Namaste, they laughed; when Hindus were washing hands and legs before entering home, they laughed; when Hindus were worshiping animals, they laughed; when Hindus were worshiping plants, trees, and forests, they laughed; when Hindus were primarily having veg diet, they laughed; when Hindus were doing Yoga, they laughed; when Hindus were worshiping God and Goddess, they laughed; when Hindus were burning the dead, they laughed; when Hindus bathed after attending a funeral, they laughed; well, guess what? Nobody is laughing now, so it is rightly said, Vedic culture in India is art of living, not a religion [44, 45]. The sacrificial fire has anthelmintic and anti-inflammatory properties. Where there are yagyas, there is definitely a shortage of collective diseases (Yajna: a holistic healing process, Yugrishi Shriram Sharma Acharya) [31, 42].

Yajna and Mantra Chanting as a Therapy  299

16.7 Applications of Yajna Therapy It is from tenacity that tremendous power is produced, which can turn the direction of the public. Gayatri is a divine discipline, which God has made accessible to us. The sages and monks have ordered us to benefit from Gayatri Sadhana on foot-step in the scriptures, even if we do not benefit from it, if we do not do spiritual practice, what else it can be called besides misfortune (Gayatri MahaVigyan, Pt. Shriram Sharma Acharya) [46, 47]. Performing Yajna is like opening a big hospital, so that the people who are not counted sick and fall ill in future get treated at their home. Difficulties on medical are of money, time, labor, anxiety, loss of work, etc. Many diseases would be saved from this, thus sacrificing even the opening of the hospital. It proves more useful (Yagya: an enlightened healing process, Yugrishi Shriram Sharma Acharya) [48, 49]. In the dark years of the house, behind the shell, stuff, etc., in the cracks of the wall and secret from the secret, and in places, the pathogens remain hidden, and they are from the sacrificial smoke. They get destroyed (Yagya: a holistic healing process, Yugrishi Shriram Sharma Acharya) [32, 40].

16.8 Conclusions The will, be of main, affects not only the conscious world but also the unconscious things. On the strength of his will, man interviews God in stone, metal, etc. By strong will, man can also affect the root, conscious, and nature. The houses where the Yajna is performed also become a kind of pilgrimage and the people whose stay there are high, well developed, and cultured (Gayatri MahaVigyan, Yugrishi Shriram Sharma Acharya) [50]. Prayag is considered the king of all pilgrimages; similarly, the Triveni of spiritual world is Gayatri (Gayatri MahaVigyan, Pt. Shriram Sharma Acharya) [33, 43]. By the given manuscript, the effect of Yajna and Mantra over Human Health is beautifully described.

Acknowledgement We would like to acknowledge with much appreciation IIT Delhi and IIT Roorkee. The guidance given by Patanjali Foundation and Ayurved Institute Dehradoon helped me to reach significant conclusions. Also, the author team expresses thankfulness to Dev Sanskriti Vishvavidyala Haridwar for helping with the research content. Furthermore, we would like to thank all the other people who were directly or indirectly involved with this research and kept on supporting us.

References 1. Acharya, D.P. and Ahmed, P.K., A Survey on Big Data Analytics. Int. J. Adv. Comput. Sci. Appl., 7, 2, 1–13, 2016.

300  The Smart Cyber Ecosystem for Sustainable Development 2. Ai, Y., Peng, M., Zhang, K., Edge Computing Technologies for Internet of Things: a primer. Digital Commun. Networks, 4, 77–86, 2018. Available: https://www.sciencedirect.com/science/ article/pii/S2352864817301335. 3. Chaturvedi, D.K., Science, Religion and Spiritual Quest, in: Edited book on Linkages between Social Service, Agriculture and Theology for the Future of Mankind, pp. 15–17, DEI Press, 2004. 4. Chaturvedi, D.K., Human Rights and Consciousness, International Seminar on Prominence of Human Rights in the Criminal Justice System (ISPUR 2012), Organized Ambedkar Chair, Dept. of Contemporary Social Studies & Law, p. 33, Dr. B.R. Ambedkar University, Agra, 2012a. 5. Chaturvedi, D.K., Relationship between Chakra Energy and Consciousness. Biomed. J. Sci. Tech. Research, 15, 3, 1–3, 2019. 6. Chaturvedi, D.K., Arora, J.K., Bhardwaj, R., Effect of meditation on Chakra Energy and Hemodynamic parameters. Int. J. Comput. Appl., 126, 12, 52–59, 2015a. 7. Chaturvedi, D.K. and Lajwanti, Correlation between Energy Distribution profile and Level of Consciousness, ShiakshkParisamvad. Int. J. Educ. SPIJJE, 4, 1, 1–9, 2014. 8. Chaturvedi, D.K. and Lajwanti, Dayalbagh Way of Life for Better Worldliness. Quest J. J. Res. Humanities Social Sci., 3, 5, 16–23, 2015. 9. Chaturvedi, D.K., Lajwanti, Chu, T.H., Kohli, H.P., Energy Distribution Profile of Human Influences the Level of Consciousness. Towards a Science of Consciousness, Arizona Conference Proceeding, Tucson, Arizona, 2012b. 10. Chaturvedi, D.K. and Arya, M., A Study of Correlation between Consciousness Level and Performance of Worker. Ind. Eng. J., 6, 8, 40–43, 2013a. 11. Chaturvedi, D.K. and Arya, M., Correlation between Human Performance and Consciousness. IEEE-International Conference on Human Computer Interaction, Saveetha School of Engineering, Saveetha University, Thandalam, Chennai, IN, India, vol. Aug. 2013, pp. 23–24, 2013b. 12. Chaturvedi, D.K. and Satsangi, R., The Correlation between Student Performance and Consciousness Level. International Conference on Advanced Computing and Communication Technologies (ICACCT™-2013), Asia Pacific Institute of Information Technology SD India, Panipat (Haryana), Souvenir, pp. 66, proc. pp. 200–203, 16 Nov. 2013. 13. Chaturvedi, D.K. and Satsangi, R., The correlation between Student Performance and Consciousness Level. Int. J. Comput. Sci. Commun. Technol., 6, 2, 936–939, 2014. 14. Gunavathi, C. and Premalatha, K., A Comparative Analysis of Swarm Intelligence Techniques for Feature Selection in Cancer Classification. Sci. World J., 14, 12, 2014. 15. Kim, K.J. and Tagkopoulos, L., Application of Machine Learning Rheumatic Disease Research. Korean J. Intern. Med., 34, 2, 2019. 16. Mahajan, P., Application of Pattern Recognition Algorithm in Health and Medicine: a review. Int. J. Eng. Comput. Sci., 5, 5, 16580–83, May 2016. 17. Marjani, M., Nasaruddin, F., Gani, A., Karim, A., Hashem, I.A.T., Siddiqa, A., Yakoob, I., Big IoT Data Analytics: Architecture, Opportunities, and Open Research Challenges. IEEE Access, 5, 5247–5261, March 2017. 18. Mathers, C.D. and Loncar, D., Projections of Global Mortality and Burden of Disease from 2002 to 2030. PLOS Med., 15, 1, 567–574, 28 Nov. 2006. Available: https://journals.plos.org/ plosmedicine/article?id=10.1371/journal.pmed.0030442. 19. Mistry, R., Tanwar, S., Tyagi, S., Kumar, N., Blockchain for 5G-Enabled IoT for Industrial Automation: A Systematic Review, Solutions, and Challenges. Mech. Syst. Signal Process., 135, 1–19, 2020. 20. Mouradian, C., Naboulsi, D., Yangui, S., Glitho, H., Morrow, M.J., Polakos, P.A., A Comprehensive Survey on Fog Computing: State-of-the-Art and Research Challenges. IEEE, 20, 16–464, Nov. 2017. Available: https://ieeexplore.ieee.org/abstract/document/8100873/authors#authors.

Yajna and Mantra Chanting as a Therapy  301 21. Nair, R.R., Agnihotra Yajna. J. Acupuncture Meridian Stud., 10, 2, 2017, 143–150, April 2017. 22. Pang, Z., Yang, G., Khedri, R., Zhang, Y.T., Introduction to the Special Section: Convergence of Automation Technology, Biomedical Engineering, and Health Informatics Toward the Healthcare 4.0. IEEE, 11, 249–259, July 2018. Available: https://ieeexplore.org/document/8421122. 23. Rastogi, R., Chaturvedi, D.K., Arora, N., Trivedi, P., Singh, P., Role and efficacy of Positive Thinking on Stress Management and Creative Problem Solving for Adolescents. Int. J. Comput. Intell. Biotechnol. Biochem. Eng. By Mantech Publ., 2, 2, 1–27, 2017. 24. Rastogi, R., Chaturvedi, D.K., Sharma, S., Bansal, A., Agrawal, A., Audio Visual EMG & GSR Biofeedback Analysis for Effect of Spiritual Techniques on Human Behaviour and Psychic Challenges. Proceedings of the 12th INDIACom; INDIACom-2018, pp. 252–258, 14th–16 March, 2018a. 25. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Sirohi, H., Singh, M., Verma, P., Singh, V., Which One is Best: Electromyography Biofeedback Efficacy Analysis on Audio, Visual and Audio-Visual Modes for Chronic TTH on Different Characteristics. proceedings of International Conference on Computational Intelligence &IoT (ICCIIoT) 2018, 14-15 December 2018b, National Institute of Technology Agartala, Tripura, India, ELSEVIER-SSRN Digital Library. 26. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Saini, H., Verma, H., Mehlyan, K., Comparative Efficacy Analysis of Electromyography and Galvanic Skin Resistance Biofeedback on Audio Mode for Chronic TTH on Various Indicators. Proceedings of International Conference on Computational Intelligence & IoT (ICCIIoT) 2018, 14-15 December 2018c, National Institute of Technology Agartala, Tripura, India, ELSEVIER-SSRN Digital Library (ISSN 1556-5068). 27. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Singh, P., Vyas, P., Statistical Analysis for Effect of Positive Thinking on Stress Management and Creative Problem Solving for Adolescents. Proceedings of the 12thINDIACom; INDIACom-2018, pp. 245–251, 14th–16 March, 2018d. 28. Rastogi, R., Chaturvedi, D.K., Arora, N., Trivedi, P., Singh, P., Vyas, P., Study on Efficacy of Electromyography and Electroencephalography Biofeedback with Mindful Meditation on Mental health of Youths. Proceedings of the 12th INDIACom; INDIACom-2018, pp. 84–89, 14th–16 March, 2018e. 29. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Saini, H., Verma, H., Mehlyan, K., Varshney, Y., Statistical Analysis of EMG and GSR Therapy on Visual Mode and SF-36 Scores for Chronic TTH. Proceedings of International Conference on 5th IEEE Uttar Pradesh Section International Conference, MMMUT Gorakhpur, 2-4 Nov. 2018. 30. Rastogi, R., Saxena, M., Sharma, S.K., Muralidharan, S., Beriwal, V.K., Singhal, P., Rastogi, M., Shrivastava, R., Evaluation of Efficacy of Yagya Therapy on T2-Diabetes Mellitus Patients, in the proceedings of The 2nd edition of International Conference on Industry Interactive Innovations in Science, Engineering and Technology (I3SET2K19) organized by JIS College of Engineering, Kalyani, West Bengal, 13-14 Dec. 2019a, https://papers.ssrn.com/sol3/papers. cfm?abstract_id=3514326. 31. Rastogi, R., Saxena, M., Gupta, U.S., Sharma, S., Chaturvedi, D.K., Singhal, P., Gupta, M., Garg, P., Gupta, M., Maheshwari, M., Yajna and Mantra Therapy Applications on Diabetic Subjects: Computational Intelligence Based Experimental Approach. proceedings of The 2nd edition of International Conference on Industry Interactive Innovations in Science, Engineering and Technology (I3SET2K19) organized by JIS College of Engineering, Kalyani, West Bengal, 13-14 Dec. 2019b, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3515800. 32. Rastogi, R., Saxena, M., Sharma, S.K., Murlidharan, S., Berival, V.K., Jaiswal, D., Sharma, A., Mishra, A., Statistical Analysis on Efficacy of Yagya Therapy for Type-2 Diabetic Mellitus Patients through Various Parameters, in: Computational Intelligence in Pattern Recognition (CIPR), Kalyani, West Bengal. Advs in Intelligent Syst., Computing, vol. 1120, A. Das, J. Nayak,

302  The Smart Cyber Ecosystem for Sustainable Development B. Naik, S. Dutta, D. Pelusi (Eds.), Computational Intelligence in Pattern, Recognition, 13-14 Dec. 2019. 33. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Yadav, V., Yadav, V., Sharma, P., Chauhan, S., Statistical Analysis of EMG & GSR Biofeedback Efficacy on Different Modes for Chronic TTH on Various Indicators. Int. J. Adv. Intell. Paradigms, Forthcoming article, X, 2019. 34. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, Gupta, M., Verma, H., Singhal, P., Singh, A., Comparative Study of Trends Observed During Different Medications by Subjects under EMG & GSR Biofeedback. ICSMSIC-2019, by MCA and IT Dept. of ABESEC, Ghaziabad. 8-9 March 2019.IJITEE, vol. 8, pp. 748–756, 2019e, https://www.ijitee.org/download/ volume-8-issue-6S/. 35. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Gupta, M., Singhal, P., Gulati, M., Statistical Analysis ofExponential and Polynomial Models of EMG & GSR Biofeedback for Correlation between Subjects’ MedicationsMovement & Medication Scores. ICSMSIC-2019, by MCA and IT Dept. of ABESEC, Ghaziabad. 8-9 March 2019.IJITEE, vol. 8, pp. 625–635, 2019f, https:// www.ijitee.org/download/volume-8-issue-6S/. 36. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Gupta, M., Saini, H., Mahelyan, K.S., Verma, H., Comparative Efficacy Analysis of Electromyography and Galvanic Skin Resistance Biofeedback on Audio Mode for Chronic TTH on Various Indicators. Int. J. Comput. Intell. IoT, 1, 1, 18–24, March 18, 2019g. Available at SSRN: https://ssrn.com/abstract=3354371. 37. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Gupta, M., Sirohi, H., Singh, M., Verma, P., Singh, V., Which one is Best: Electromyography Biofeedback, Efficacy Analysis on Audio, Visual and Audio-Visual Modes for Chronic TTH on Different Characteristics. Int. J. Comput. Intell. IoT, 1, 1, 25–31, March 18, 2019h. Available at SSRN: https://ssrn.com/abstract=3354375. 38. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Gupta, M., Yadav, V., Chauhan, S., Sharma, P., Chronic TTH Analysis by EMG & GSR Biofeedback on Various Modes and Various Medical Symptoms Using IoT, in: volumes In series: Advances in ubiquitous sensing applications for healthcare, Book-Big Data Analytics for Intelligent Healthcare Management, Chapter 5, pp. 87–149, 2019, Academic Press, https://doi.org/10.1016/B978-0-12-818146-1.00005-2. 39. Rastogi, R., Chaturvedi, D.K., Sharma, P., Yadav, V., Chauhan, S., Gulati, M., Gupta, M., Singhal, P., Statistical Resultant Analysis of Psychosomatic Survey on Various Human Personality Indicators: Statistical Survey to Map Stress and Mental Health, in: Handbook of Research on Learning in the Age of Transhumanism, S. Sisman-Ugur and G. Kurubacak (Eds.), pp. 363–383, IGI Global, Hershey, PA, 2019. 40. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Trivedi, P., Singh, A., Sharma, A., Singh, A., Intelligent Analysis for Personality Detection on Various Indicators by Clinical Reliable Psychological TTH and Stress Surveys, in: proceedings of International Conference on Computational Intelligence in Pattern Recognition, pp. 127–144, 19th-20th January 2019k, (CIPR 2019) at Indian Institute of Engineering Science and Technology, Shibpur, Springer Advances in Intelligent Systems and Computing (AISC) Series. 41. Rastogi, R., Chaturvedi, D.K., Gupta, M., Singhal, P., Intelligent Mental Health Analyzer by Biofeedback: App and Analysis, in: Handbook of Research on Optimizing Healthcare Management Techniques, Wickramasinghe, N. (Ed.), 2020. 42. Rastogi, R., Chaturvedi, D.K., Verma, H., Mishra, Y., Gupta, M., Identifying Better? Analytical Trends to Check Subjects’ Medications Using Biofeedback Therapies, IGL Global. Int. J. Appl. Res. Public Health Manage. (IJARPHM), 5, 1, 2, 14–31, 2020. 43. Rastogi, R., Gupta, M., Chaturvedi, D.K., Efficacy of Study for Correlation of TTH vs Age and Gender Factors using EMG Biofeedback Technique. Int. J. Appl. Res. Public Health Manage. (IJARPHM), 5, 1, 4, 49–66, 2020.

Yajna and Mantra Chanting as a Therapy  303 44. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Gupta, M., Verma, H., Saini, H., An Optimized Biofeedback EMG and GSR Biofeedback Therapy for Chronic TTH on SF-36 Scores of Different MMBD Modes on Various Medical Symptoms, in: Studies Comp. Intelligence, Vol. 841, Hybrid Machine Intelligence for Medical Image Analysis, S. Bhattacharya et al. (Eds.), vol. 8, 2020c, 978-981-13-8929-0, 468690_1_En, 2020c, https:// doi.org/10.1007/978-981-13-8930-6_8. 45. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Trivedi, P., Singh, A.K., Sharma, A.K., Singh, A., Intelligent Personality Analysis on Indicators in IoT-MMBD Enabled Environment, in: Multimedia Big Data Computing for IoT Applications: Concepts, Paradigms, and Solutions, S. Tanwar, S. Tyagi, N. Kumar (Eds.), pp. 185–215, Springer Nature, Singapore, 2020d, https://doi. org/10.1007/978-981-13-8759-3_7. 46. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Trivedi, P., Gupta, M., Singhal, P., Gulati, M., MM Big Data Applications: Statistical Resultant Analysis of Psychosomatic Survey on Various Human Personality Indicators. Proceedings of Second International Conference on Computational Intelligence 2018, 2020, https://doi.org/10.1007/978-981-13-8222-2_25. 47. Rastogi, R., Chaturvedi, D.K., Gupta, M., Singhal, P., Surveillance of Type –I & II Diabetic Subjects on Physical Characteristics: IoT and Big Data Perspective in Healthcare @NCR, India, in: Chapter No. 23, Internet of Things (IoT), Concept and Applications, Alam, M., Shakil, K.A., Khan, S. (Eds.), 10 May 2020. 48. Rastogi, R., Chaturvedi, D.K., Gupta, M., Sirohi, H., Gulati, M., Pratyusha, Analytical Observations between Subjects’ Medications Movement & Medication Scores Correlation Based on their Gender and Age Using GSR Biofeedback: Intelligent Application in Healthcare, in: Pattern Recognition Applications in Engineering, M.A. Vejar and F. Pozo (Eds.), pp. 229–257, IGI Global, Hershey, PA, 2020, |Pages: 229–257. 49. Rastogi, R., Chaturvedi, D.K., Gupta, M., Sirohi, H., Gulati, M., Pratyusha, Analytical Observations Between Subjects’ Medications Movement and Medication Scores Correlation Based on Their Gender and Age Using GSR Biofeedback: Intelligent Application in Healthcare, in: Pattern Recognition Applications in Engineering, D. Burgos, M. Vejar, F. Pozo (Eds.), pp. 229–257, IGI Global, Hershey, PA, 2020. 50. Rastogi, R., Chaturvedi, D.K., Gupta, M., Exhibiting App and Analysis for Biofeedback Based Mental Health Analyzer, in: as chapter 15 in Handbook of Research on Advancements of Artificial Intelligence in Healthcare Engineering, p. 300, 2020. 51. Rastogi, R., Chaturvedi, D.K., Gupta, M., Computational Approach for Personality Detection on Attributes: An IoT-MMBD Enabled Environment, in: as Chapter 16 in Handbook of Research on Advancements of Artificial Intelligence in Healthcare Engineering, p. 300, 2020. 52. Rastogi, R., Chaturvedi, D.K., Gupta, &, M., Tension Type Headache: IoT and fog Applications in Healthcare Using Different Biofeedback, in: chapter 17 of Handbook of Research on Advancements of Artificial Intelligence in Healthcare Engineering, p. 300, 2020. 53. Rastogi, R. and Chaturvedi, D.K., Tension Type Headache: IOT Applications to Cure TTH Using Different Biofeedback: A statistical Approach in Healthcare, in: In Simon George Taukeni (University of Namibia, Namibia) (Eds.) Biopsychosocial Perspectives and Practices for Addressing Communicable and Non-Communicable Diseases, Ch. 10, IGI Global, Hershey, PA, 2020. 54. Rastogi, R., Saxena, M., Maheshwari, M., Garg, P., Gupta, M., Shrivastava, R., Rastogi, M., Gupta, H., Yajna and Mantra Science Bringing Health and Comfort to Indo-Asian Public: A Healthcare 4.0 Approach and Computational Study, in: Machine Learning with Healthcare Perspective. Learning and Analytics in Intelligent Systems, vol. 13, V. Jain and J. Chatterjee (Eds.), pp. 357–390, Springer, Cham, Springer Nature, Switzerland, AG 2020, 2020, https:// link.springer.com/chapter/10.1007%2F978-3-030-40850-3_15.

304  The Smart Cyber Ecosystem for Sustainable Development 55. Rastogi, R., Chaturvedi, D.K., Satya, S., Arora, N., Intelligent Heart Disease Prediction on Physical and Mental Parameters: A ML Based IoT and Big Data Application and Analysis, in: Machine Learning with Healthcare Perspective. Learning and Analytics in Intelligent Systems, vol. 13, V. Jain and J. Chatterjee (Eds.), pp. 199–236, Springer, Cham, Springer Nature, Switzerland 2020. 56. Rastogi, R., Chaturvedi, D.K., Singhal, P., Gupta, M., Investigating Diabetic Subjects on Their Correlation with TTH and CAD: A Statistical Approach on Experimental Results, in: Opportunities and Challenges in Digital Healthcare Innovation, Sandhu, Dr. K. (Ed.), 2020. 57. Rastogi, R., Chaturvedi, D.K., Singhal, P., Gupta, M., Investigating Correlation of Tension Type Headache and Diabetes: IoT Perspective in Healthcare, in: IoTHT: Internet of Things for Healthcare Technologies, Chakerborty, C. (Ed.), Springer Nature, Singapore, 2020. 58. Saxena, M., Sengupta, B., Pandya, P., A study of the Impact of Yagya on Indoor Microbial Environments. Indian J. Air Pollut. Control, 7, 1, 6 – 15, 2007. 59. Saxena, M., Sengupta, B., Pandya, P., Comparative Studies of Yagya vs. Non-Yagya Microbial Environments. Indian J. Air Pollut. Control, VII, 1, 16 – 24, 2007. 60. Saxena, M., Sengupta, B., Pandya, P., Effect of Yagya on the Gaseous Pollutants. Indian J. Air Pollut. Control, 7, 2, 11–15, Sept. 2007. 61. Saxena, M., Sengupta, B., Pandya, P., Controlling the Microflora in Outdoor Environment: Effect of Yagya. Indian J. Air Pollut. Control, 8, 2, 30 – 36, Sept. 2008. 62. Saxena, M., Kumar, B., Matharu, S., Impact of Yagya on Particulate Matters. Interdiscip. J. Yagya Res., 1, 1, 01–08, Oct. 2018. 63. Saxena, M., Sharma, S.K., Muralidharan, S., Beriwal, V., Rastogi, R., Singhal, P., Sharma, V., Sangam, U., Statistical Analysis Of Efficacy Of Yagya Therapy On Type-2 Diabetic Mellitus Patients on Various Parameters. Proceedings Of 2nd International Conference on Computational Intelligence In Pattern Recognition (CIPR – 2020), Institute Of Engineering And Management, Kolkata, West Bengal, India, 4th–5th January, 2020. 64. Sharma, S.R. and Kunj, S., The Integrated Science of Yagna. DSIIJ, 01, 04, 14, 16–17, 2001. 65. Thakur, G.S., Yajña-A Vedic Traditional Technique for Empirical and Transcendental and Achievement. Indian Streams Res. J., 04, 5, May 2014.

Key Terms and Definitions Yajna: Yajna literally means “sacrifice, devotion, worship, offering” and refers in Hinduism to any ritual done in front of a sacred fire, often with mantras. Yajna has been a Vedic tradition, described in a layer of Vedic literature called Brahmanas, as well as Yajurveda. The tradition has evolved from offering oblations and libations into sacred fire to symbolic offerings in the presence of sacred fire (Agni). Mantra: A mantra is a sacred utterance, a numinous sound, a syllable, word or phonemes, or group of words in Sanskrit believed by practitioners to have psychological and/or spiritual powers. Some mantras have a syntactic structure and literal meaning, while others do not. Jap: Jap is the meditative repetition of a mantra or a divine name. It is a practice found in Hinduism, Jainism, Sikhism, Buddhism, and Shintoism. The mantra or name may be spoken softly, enough for the practitioner to hear it, or it may be spoken within the reciter’s mind. Jap may be performed while sitting in

Yajna and Mantra Chanting as a Therapy  305 a meditation posture, while performing other activities, or as part of formal worship in group settings. Ayurveda: Ayurveda system of medicine with historical roots in the Indian subcontinent. Globalized and modernized practices derived from Ayurveda traditions are a type of alternative medicine. In countries beyond India, Ayurvedic therapies and practices have been integrated in general wellness applications and in some cases in medical use. The main classical Ayurveda texts begin with accounts of the transmission of medical knowledge from the Gods to sages, and then to human physicians. In Sushruta Samhita (Sushruta’s Compendium), Sushruta wrote Dhanvantari, Hindu god of Ayurveda. Sanskrit: Sanskrit is an Indo-Aryan language of the ancient Indian subcontinent with a 3,500-year history. It is the primary liturgical language of Hinduism and the predominant language of most works of Hindu philosophy as well as some of the principal texts of Buddhism and Jainism. Sanskrit, in its variants and numerous dialects, was the lingua franca of ancient and medieval India. In the early 1st millennium AD, along with Buddhism and Hinduism, Sanskrit migrated to Southeast Asia, parts of East Asia, and Central Asia, emerging as a language of high culture and of local ruling elites in these regions. Vedic: The Vedic period or Vedic age (c. 1500 to c. 500 BCE) is the period in the history of the northern Indian subcontinent between the end of the urban Indus Valley Civilization and a second urbanization which began in the central Indo-Gangetic Plain c. 600 BCE. It gets its name from the Vedas, which are liturgical texts containing details of life during this period that have been interpreted to be historical and constitute the primary sources for understanding the period. These documents, alongside the corresponding archaeological record, allow for the evolution of the Vedic culture to be traced and inferred.

17 Extraction of Depression Symptoms From Social Networks Bhavna Chilwal* and Amit Kumar Mishra Department of Computer Science and Engineering, DIT University, Dehradun, Uttarakhand, India

Abstract

Health is a priority for humans. Every person wants to be healthy and fit during his life span. But the inner body cycle is not under the control. There are so many health issues and diseases prevail in this world which affects the working of the human biological system. The medical field has solutions for most of the health problems. As per the growing advancement in technology and living standards, there is also an increase in different health problems. It is believed that the brain in the human body is one of the most important parts and this is the organ that is responsible for the effectiveness of any human. But nowadays, the most common diseases are mental health diseases. According to the World Health Organization (WHO), mental illness has a wide range of problems and these problems have a different combination of symptoms like confused emotions, abnormal thoughts, and fluctuation in behavior. There are many mental disorders by which people are suffering like intellectual disabilities, schizophrenia, and depression. The social network is the irreplaceable thing in today’s life. This is the platform where people share their feelings, views, thoughts with other people by making online conversations. Nowadays, different techniques are focusing on the data which is obtained daily on the internet and it increases day by day. Like many business, companies monitor people’s view about their products. Similarly, these data could be used in medical field for making a data set about people’s view for specific diseases or treatments. This chapter focuses on the extraction of depression symptoms or data related to depression disorder on social media platforms. Here, we also discussed the extraction process which involves data mining, data cleaning, and analysis. Some advance technologies like sentiment analysis, emotional analysis, and behavioral analysis are also defined here for the collection of more robust symptoms for depression disorder. Keywords:  Depression, data mining, sentiment analysis, behavior analysis, social network

17.1 Introduction According to WHO, depression is the most common mental disorder among humans. Around 264 million people are suffering from this disorder. It is characterized when a person finds no interest and pleasure in rewarding activities and has constant sadness, which continuously leads to tiredness, insomnia, poor concentration, and disturbed appetite. The *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (307–322) © 2021 Scrivener Publishing LLC

307

308  The Smart Cyber Ecosystem for Sustainable Development depression has long-lasting effects and disturbs a person’s living habits and the ability to perform any work. Many factors cause this disorder like difficult social life and complex biological and psychological interactions. In some cases, it may be due to worse life events in someone’s life like in childhood or adulthood and, sometimes, due to bad working life, unemployment, and bad relationships. Its worst face is sometimes leads to suicide [14, 15]. As per WHO nearly 8 lac people attempted suicide every year due to mental disorders. It has been noted that more women are likely to suffer from depression as compared to men. As per the severity level of this disorder, a combination of symptoms is distinguished for categories like mild, moderate, and severe, and these categories are used to differentiate between types of depressive disorders which prevail nowadays. Figure 17.1 shows the types of depression disorders. 1. Major Depression Disorder: While suffering from this illness, the person is depressed most of the time in a day, and there is the loss of interest in any activity, bad sleep, feeling guilty, restless nature, and suicidal thoughts; these symptoms are occurring for more than 2 weeks, and then, doctors suggest therapy and medications. Sometimes, in worst situations, the doctors suggest electroconvulsive therapy (ECT) and repetitive transcranial magnetic stimulation (rTMS), which uses a kind of magnet for stimulation of brain activities to control moods of a person and he/she could work better. 2. Persistent Depressive Disorder: For this disorder, the duration is longer than 2 years. For this disorder, the combination of symptoms is different like eating disorder, feeling helpless, disturb sleep, low self-confidence, and difficulty to make any decision. Psychotherapy and medication are helpful for this. 3. Bipolar Disorder: This is also known as manic depression. Here, the person has mood swing episodes like sometimes the energy of a person is extremely high called “UP” mood, and sometimes, the same is feeling very low and depressed. Doctors suggest medications and mood stabilizers for this. Also, psychotherapy helps here. 4. Seasonal Effective Disorder: The abbreviation for this illness is SAD which generally occurs in the winter season as a person feels major depression when it gets less sunlight and the days are short. For SAD, doctors suggest antidepressant and light therapy. It typically goes away in the spring and summer season. Major Depression

Bipolar Disorder

Psychotic Depression

Depression Disorder

Persistent Depressive Disorder

Figure 17.1  Types of depression.

Seasonal Effective Disorder

Extraction of Depression Symptoms  309 5. Psychotic Depression: This disorder has psychotic symptoms like the delusions where the person makes false beliefs, has hallucination that means he sees and hears things which are not present, and has paranoia that means every time he feels that someone is going to harm him. For this, psychotherapy and ECT are helpful. So, these are the major types of depression that can cause the worst effect in the patient’s life. Sometimes, these disorders could lead to suicide and major mental problems if they are not taken care of at the right time. The depression also has an interrelationship with physical health, and it could cause by any other physical problem like chronic health issues, which leads to depression. Depression problem is not implied weakness or a negative personality. This disorder is a major health problem nowadays which has its treatments and medications. It is common to feel low sometimes in daily life but if this prevails on a longer period, then it is time to see a doctor because; otherwise, it could be chronic and suicidal.

17.1.1 Diagnosis and Treatments Good quality diagnoses and treatments are available for depression disorder categories. Different psychological treatments like CBT, i.e., cognitive behavioral therapy, and IPT, i.e., interpersonal psychotherapy and behavioral activations, are provided by healthcare providers. The combination of symptoms is applied to detect among the type of disorder a person is suffering from. For a collection of symptoms, questionnaires are prepared by the expert psychologist according to different situations of day-to-day life, so that the patients could give their responses in the best effective manner. Sometimes, in moderate and severe conditions, the psychologists prefer face-to-face or group interview questionnaires between the therapist and the patient. The interview process is very intense as it captures every single movement of the patient to make the best possible response and make the correct combination of symptoms. Nowadays, ECT and rTMS stimulation device is used for getting each brain activity and form symptoms. Antidepressant medications are also available to treat illness. Although, the advancement in diagnosis and treatment measures is still between 70% and 85% of the population in low and middle-income countries do not have effective treatments for their disorders. The reasons are inadequate resources, lack of good health providers, and therapists. Sometimes, because of using incorrect methods to collect the symptoms and inaccurate assessments, it leads to the wrong diagnosis of depression. Also, these methods are tiresome and expensive, because of which people try to avoid it. Also, in questionnaires, the proper responses are not collected because of the lengthy questions which people fill irresponsibly and sometimes those who do not have the illness are misdiagnosed. The depression disorder is increasing globally at a high rate, and according to the National Institute of Mental Health, more than 15 million adults and 2 out of every 100 children are suffering from depression disorder [4]. It is not possible to take face to face interviews and collect questionnaire responses for such big numbers. It becomes difficult to get the symptoms for such a large population. The researchers are working continuously for the betterment of depressive disorder. Worldwide researches have been done for detections and collection of symptoms for this disorder, the medical depression diagnosis is a very complex and long-term process.

310  The Smart Cyber Ecosystem for Sustainable Development Sometimes, the interview and questionnaires are not enough to collect all symptoms for a specific disorder. The researchers are working hard to find some methods which discover more and accurate symptoms specific to a particular type of disorder. For this, more focus goes toward data mining techniques.

17.2 Data Mining in Healthcare The data mining methods implementation in the health sector is an emerging field for researchers as it provides great help in understanding the diseases and their classification, and also, this field has a great impact on the mental health sector. Data mining is a technology that looks for hidden patterns, which are valid and potential among the huge data available. This also involves discovering unknown information and relationships among datasets. This field is a combination of different scientific and research fields like AI, machine learning, database technology, and statistics. The data mining is also known as knowledge and information extraction, information harvesting, pattern analysis, etc. This field has been used in various research techniques in computer science like classification, clustering, regression, prediction, and association rules. Data mining has different types to extract information from today’s data world. Text mining is among one of the types. Text mining is the information and data extractions from any available written resources. These resources could be emails, social media comments, blogs, books, and articles. This mining is a kind of text analysis that involves the process of extracting high-quality information from large texts.

17.2.1 Text Mining Figure 17.2 illustrates the text mining process which involves the structuring of text which considers being input and then extracts the useful patterns from that text, and finally, interpretation and evaluation are performed to get the output. The main goal of text mining is to turn the text into some useful information. The researchers nowadays frequently use text mining as today’s world is revolving around data [1, 3]. Every day, more than 2.5 quintillion bytes of data are created in which 70% data is the internet data. Now, the question is what is internet data? Internet data is the service provided by telecom companies to its users for using internet services at different prices. Most of the end-user uses this service for accessing social media sites which are also called social network sites. The text analyses have

Heap of Text

Figure 17.2  Text mining.

Useful Text

Extraction of Depression Symptoms  311 information extraction and lexical coding to the recognition of patterns and study word frequency including association, predictive, and visualization analysis [7]. The text mining also uses the new application which nowadays becomes the wide area of research that is NLP, i.e., Natural Language Processing.

17.3 Social Network Sites Social media is very huge; as per reports, 550 new users per minute join social media platforms, i.e., 300 million new users every year. Also, since 2013, 58% of tweets have increased each minute in tweeter platform. Instagram users upload daily more than 100 million photos and more than 70,000 million posts in every single minute. Average users spend 1 hour per day of internet on Facebook. On Facebook platform, around 500,000 comments and 290,000 statuses get posted in each minute. Over 100 million text messages are sent on the messaging app in every single minute [8]. Worldwide, around 2 trillion Google searches happen. So, every minute in a day generates loads of data. These social media sites are platforms where people communicate with each other through comments, posts, status, and blogs. When such a large volume of data is generated, it becomes very difficult to analyze and evaluate it so that text mining technology is used. Figure 17.3 shows the pie chart of social networking sites according to their active users. Nowadays, social networking sites become the most important part of people’s life where they share their thoughts, their life events, opinions, and their well beings. They feel these platforms so comfortable that they communicate about their work life, love life, personal problems, their career interests, and other forms of expression. It is noticed that the written posts and comments on social network somehow show the mood and the behavior of the person at that moment. Sometimes, the data of these posts help to specify the event discussed between the users like they are happy or sad about the event. So, by seeing the advancement and the involvement of these platforms, the researchers use this network data for exaction of information about the behavior of different persons, their moods, and their expression by just analyzing their messages and comments by using text mining applications. The advancement in the social network and text mining techniques could also be used for mental illness. NO. OF ACTIVE USERS IN MILLIONS Youtube

Facebook

Twitter

Instagram

Whatsapp

Figure 17.3  Pie chart of social networking sites as per the number of users.

312  The Smart Cyber Ecosystem for Sustainable Development As depression is the most common problem nowadays and it has challenges while detection as the depression symptoms may vary from person to person because of differences in behavior and personality. It becomes difficult to get a clinical report for each person of such a large population; moreover, these records have restrictions on gender and age. So, to help the medical team working on depression disorder, the text mining tools are applied on social media data to extract and specify symptoms for depression because medical data is not enough. Social media sites like Twitter, Facebook, Instagram, Snapchat, Messenger Applications, and Blogs generated countless data every day. But for data extraction focusing on mental illness, there is a requirement of advanced tools like NLP, machine learning, social network analysis, and sentiment analysis.

17.4 Symptom Extraction Tool NLP is a computational technique that analyzes the natural speech and language. Social media has data that comprises natural language used by people in their day-to-day life. So, to analyze such large data, NLP helps to establish interaction between the computer and human natural language. Similarly, people share their thoughts and opinions about mental illness in natural language by texting. So, to extract the symptoms or specific behavior of any person, the NLP is applied for deriving the conclusions and detecting the symptoms from the text. But for applying the NLP, the computers should be smart enough to extract the required word which will identify the symptoms of depressive disorder. To make the computer systems efficient for the extraction of symptoms for depression needs some techniques and methods. Here, in this chapter, the symptom extraction techniques are discussed in brief. Figure 17.4 illustrates the process of extraction of symptoms.

Large Data

Data Collection

Data Preprocessing

Data from Social Media

Data gets collected related to depression symptoms

Data Cleaning and transformation

Word Frequency Data Analysis

Word Embedding Word Clustering

Representation

Figure 17.4  Symptom extraction process.

Extraction of Depression Symptoms  313

17.4.1 Data Collection The data collection process is performed by recording the social media active user’s inputs who have discussed the mental health illness. Every minute, the social media platforms have millions of comments and posts. Facebook and Twitter are the most widely used social media site which has millions of active users daily. So, to gather the data related to depression, there will be continuous monitoring of each posted or tweeted text which has the keywords like depression, stress, irritation, mood off, less sleep, sad, fear, help, psycho, eating problems, tired, helpless, disappointment, uninterested, boring, bad temper, heavy heart, crying, low, tension, dullness, rejected, unemployed, gloom, sorrow, trouble, blue, worry, misery, distress, hopeless, and also those words related to these words which help to find the posts related to depression. Also, the data could be collected by the tweets and posts done by the mental health experts or the people posted their messages on these mental health pages or profiles owned by the health providers. One more method for data collection is to follow the websites and blogs which are aimed at depression and mental illness. The monitoring is done for the regular users of these blogs and websites who post their comments and queries related to mental problems. The collection required proper monitoring channels because of the complexity of data available on social networking sites. As the data gets collected related to the depression disorder, it is not all because the collected data could be a combination of different natural language words and phrases which need processing for implementation. So, the next method after collection is data processing.

17.4.2 Data Processing The role of data processing is to convert the raw, noisy, and unstructured data into an efficient and useful form. The social media data is very raw and redundant. Sometimes, the collected data phrase has less information related to the depression field and sometimes the post comprises of stop words and non-words which are not helpful for the extraction of symptoms. So, to avoid these problems, data processing is done after data collection. The processing techniques involve the following: a. Data Cleaning b. Data Transformation c. Data Reduction Sometimes, data has many missing and irrelevant parts which are not enough to provide any specific information about depression. This data could be meaningless and full of errors, so to handle such data, data processing techniques are used here to transforming unstructured raw data into useful information, and it is done by the help of machine learning methods or NLP Toolkit (NLTK). Generally, the NLTK methods help to remove special characters, stop words, non-words, and missing and meaningless words. After the cleaning and preprocessing of the data related to depression symptoms, we need to analyze them by doing data analysis.

314  The Smart Cyber Ecosystem for Sustainable Development

17.4.3 Data Analysis The data analysis is the process of modeling the data in the form so that new information could be discovered and the data could be used for the decision making process. In text mining, analyzing data means the pattern in which data is received, what is the frequency of words, counting the word occurrences related to depression [24]. As it is believed in text mining that the most commonly used words in the text are important in natural language. So, proper monitoring and analysis is performed to capture the frequently used words which are synonyms of depression symptom words and universal dataset will be maintained for these important words whose count is very high in natural language posts in social media like “anxiety”, “distress”, and so on. So, word count or frequency plays an important role in analyzing the text related to a specific field like here the goal is to find important words that are related to depression symptoms. Figure 17.5 represents the few words related to depression and a bar chart of their frequency. Now, as the important words or symptoms related words are picked, it is time to make a good vocabulary or word dictionary especially for the depression symptoms. The process of binding the words in one vector or vocabulary format is known as embedding. The word embedding is the technique in NLP, in which the words of similar meaning will have similar representation. Figure 17.6 represents the word embedding process, that is, each word is mapped with one specific vector. Word embedding is a kind technique which generally represents words in predefined real-valued vectors space. Here, the mapping is performed between the words and the vectors. The vector values resemble the neural network learning ways. The word embedding needs clean and preprocessed text in which each word is one-hot encoded. The embedding layer is formed by using a neural network and the vector size space is defined as the size of the model. This way for depression symptoms, the words get embedded in a similar type of representation from such a large volume of data. During the extraction process, word embedding plays an important role because by this we can find all the similar depression-related words in the same place or same format for future reference.

Words Frequency

6 5 4 3 2 1 0

curious Depression distress

Fear

low

Restless Sleepless Stress

Figure 17.5  Word frequency for depression related words.

Tension

Tired

Unhappy

Extraction of Depression Symptoms  315

[V1] Depression Stress

[V2]

Tension

[V3]

Low

[V4]

Unique Words

Vectors

Figure 17.6  Word embedding.

The machine learning or neural network models are used to train the systems for word embedding process. Word2Vec is one of the common models for word embedding training. It is a statistical method for learning standalone web embedding for the systems. Two other learning methods were also included with Word2Vec that is Continuous Skip-gram model and Continuous Bag-of-Words model. For depression symptoms, word embedding could be learned by text data which has millions of words and training is also performed by focusing on standalone learning methods or by learning jointly especially for the depression model. There is one more data analysis process that could make the extraction process more advance and accurate focusing on the depression, and the process is called word clustering. Figure 17.7 illustrates the different clusters for different symptoms of depression. The clustering is the more advanced extraction techniques of required data from the text available. It comprises dividing the words into certain clusters or groups of the same category. Every cluster contains words that are related to each other syntactically or semantically. Here, for the extraction of depression symptoms, the word clustering is applied which partition the set of the word into groups and these groups have their characteristics related to depression disorder and consist of words that are similar in meaning concerning the particular cluster. For applying the clustering, K-means algorithm is used generally for training the system to learn word clustering. The clustering technique is a type of unsupervised learning algorithm. By applying the word clustering method, the relationship will we establish between the words and discover useful symptoms for depressive disorder [22, 23].

distress

Sleeping Disorder

uneasiness

restless

Eating Disorder

insomia

Fear tired

Figure 17.7  Word clustering.

unrest

fat

trouble curious

igu

e

Anxiety

316  The Smart Cyber Ecosystem for Sustainable Development The extraction process which is discussed above is using the text mining method on the social network data and the main focus is on the comment, post, and messages by which people communicate with one another and those post will be monitored which contains any text or word related to word depression to make the helpful dataset in natural language because the medical data and terms are not used in day-to-day discussion. Depression has different classes of emotions and behavior as well [25]. So, by using the extraction process, the natural language depression symptoms or words could be formed and it will help doctors and psychologists to form more records, symptoms, and behavioral characteristics for this disorder. But, except the extraction of useful text, it is not enough if the context and the sentiments are covered. So, to make the process more reliable and efficient, the sentimental analysis is also used to get the real meaning and expression of the text.

17.5 Sentiment Analysis

Ne 18 utra % l

The sentiment analysis is a branch of data mining, which is also known as contextual mining of the text available and extracts subjective opinion and information of any product, event, personality, and characteristics, and social sentiment by monitoring conversations occurred in social media platforms. With the advancement of deep learning, the algorithms become capable enough to analyze the text [16]. It is a text classification technique which analyzes the messages or post during online conversation and tells whether the conversation is positive, negative, and neutral in context. In NLP, it is also known as opinion mining as it tries to extract the opinion from the texts. Social media is an important part of a person’s life where they reflect their real and natural life. Sometimes, people share their thoughts and views which have a sentiment it could be happiness or sadness [2, 6, 9]. So, by sentiment analysis, the main focus will be on the negative sentimental text and try to extract those sentiments which are related to the depression symptoms by using above extraction process, as by analyzing the sentiments and context will help and give an accuracy of the extraction process that the data collected is only focused to depression disorder by words as well as emotions also. Figure 17.8 shows that the sentiment analysis classifies the text document into three categories. The sentiment analysis also follows a process to check whether the conversation is positive or negative. Firstly, it breaks each text into parts like phrases, sentences, and parts of

Un

ha 39 ppy %

Figure 17.8  Sentiment analysis of text.

Hap p 43% y

Extraction of Depression Symptoms  317 speech. Then, identify those sentiment bearing words and phrases according to that assign score or label to text whether it is a positive mood context or negative mood context. Form a category and put all negative context texts in it. After the collection of negative or unhappy sentiments text, we apply the extraction process discuss in the above section, i.e., by identifying depression-related words in those negative texts [11]. The data extraction and sentiment analysis required different machine learning, artificial intelligence, data mining, and NLP algorithms for training and learning of a system. There are lots of algorithms supervised and unsupervised algorithms available [27, 28]. Table 17.1 contains the list of all machine learning supervised and unsupervised algorithms which are used nowadays for text mining purposes by different researchers. The basic process of extraction will be implemented in the machine by using the above algorithms [12]. These algorithms are clustering algorithms which will be helpful in word embedding and clustering methods for depression symptoms. There are some specific methods for sentiment analysis also which will help to implement the process of contextual mining in machines. Table 17.2 contains all NLP and machine learning algorithms and methods that are widely used by researchers these days for sentiment analysis of any text document. These approaches are used in the business sector by companies for analyzing their products review written by the customers in online platforms. There should also be noted that, sometimes, the person does not use depression keywords, and in that case, it becomes difficult to detect the symptoms and the extraction process is not enough. So, for those cases, it is important to implement some other types of sentiment analysis that are emotion analysis and behavior analysis.

Table 17.1  List of text mining algorithms. S. no.

Algorithms for text mining processes

1.

K- Means

2.

Support Vector Machine

3.

Apriori Algorithm

4.

K- nearest neighbors

5.

C4.5 Algorithm

6.

Naïve Bayes Algorithm

7.

Cart Algorithm

8.

Adaboost Algorithm

9.

Paper Rank Algorithm

10.

FP-Growth Algorithm

318  The Smart Cyber Ecosystem for Sustainable Development Table 17.2  Sentiment analysis algorithms and approaches. S. no.

Algorithms for sentiment analysis and NLP approaches

1.

Rule-Based Approach a. Stemming b. Tokenization c. Part of speech tagging d. Parsing e. Lexicon analysis

2.

Automatic Approach a. Linear Regression b. Naive Bayes c. Support Vector Machines

17.5.1 Emotion Analysis Emotion analysis is the process of recognizing human emotions. This section discusses that the human-written text in social media platforms shows the emotion that the person is carrying during the writing of that text, and emotion detection contains some specific vocabulary of words that are used to express emotions [21]. The extraction of emotions can be done in two forms, i.e., written texts and online conversations. The online conversations in social platforms extract opinions and the mental states of the participants involved, and the person who might have the depressive disorder is generally used negative emotion words; also, the memory and brain patterns are different for the depressive person from the normal person [12]. Brain memory generally tends to corporate with a person’s current emotion and mood, and the depressive person tends to recall negative events or bad events from the past and feels sad and depressed. Even in social media, people express their emotions by using emoticons for happy, sad, like, and love [17]. So, the extraction of text also takes care of the emotions of a person because it is not necessary that depression keyword is always there; the emotions also matter for the collection of symptoms for depression disorder [13].

17.5.2 Behavioral Analysis Behavioral analysis is a science that studies human behavior and comprises understanding, describing, and predicting change in behavior. This analysis helps to identify the behavior of a person who involves in an online conversation. The depressive person generally has a change in behavior and mood swing problem [5, 25]. Also, this analysis will help to detect those persons who have behavior as a silent person and he usually not get interested in joyful moments, but it does not mean that he/she is depressive in nature. So, for those cases, this analysis will also extract only those symptoms which are specific to the disorder [18]. This behavioral analysis, nowadays, becomes one of the widely researched topics, and it helps in the treatment of behavioral problems [26]. It is reported by doctors and psychologists that people who are suffering from depression disorder also have behavioral change problems; social network sites have plenty of options in which people show their real behavior and also discussed their behavioral change with their close ones [10]. During the extraction of symptoms, the proper monitoring will be done for behavioral problems

Extraction of Depression Symptoms  319 Table 17.3  Different characteristics of sentiment, emotion, and behavioral analysis. Characteristics

Sentiment analysis

Emotion analysis

Behavior analysis

Polarity

23%

47%

30%

Phrase

I love India - Positive I hate black color Negative

Emoticons:

“I am Proud of You” - Motivating “I don’t like a party” Introvert

- Happy - Like

Attribute

Text Analysis

Self-Awareness, Empathy, Social Skills

Diligence, Ethics, Honesty

Entity

Document, Text, Paragraph

Text, Reactions, Audio

Text, Response, Audio, Video

Recommended Systems

Rule-Based System (NLP)

Facial Detection APIs

Behavioral Analysis and Testing System (BATS)

and also for emotional changes. Table 17.3 shows the difference between the sentiment, emotion, and behavioral analysis as per today’s technological advancement.

17.6 Conclusion This chapter mainly focuses on the worldwide prevailing mental illness, i.e., depression disorder. This mental problem has its types and their symptoms for detection. Researchers are working to help medical field professionals by integrating advanced computing techniques with medicine. Here, text mining technique extracts depression symptoms from the text available, and the main repository for data and text nowadays is social networking sites where people share their thoughts, happiness, opinions, personal life, and health problems to each other through different platforms available under social media. The main focus will remain on the text, messages, and post which have any specific words related to the depression and its symptoms. Also, sentiment analysis is discussed because while extraction of the depression-related keywords is not enough, and there is a need to check for the context and meaning of the expression. For sentiment analysis, some algorithms are listed which will help to train machines for classifying the text of positive and negative purpose [19, 20]. Also, the sentiment analysis is not enough as it only classifies the text in positive and negative class, it is also necessary to find the emotions which a person carries while writing any online text. For that, emotion analysis is incorporated in the extraction process; also, to avoid considering wrong data, behavioral analysis is also implemented as it will help to monitor the change in behaviors for people who are suffering from depression disorder. So, the social networking data is taken into consideration because, here, different people with different personality express their views and thoughts, as in clinical records, the symptoms for the detection of depression disease are not enough for such a large population

320  The Smart Cyber Ecosystem for Sustainable Development with diverse personality. That is why social network data is used to extract more precise depression disorder and their respective symptoms for different personalities and situations. This extraction process will help the health providers by providing a variety of records corresponding to the symptoms related to depression disorder and help the people who are suffering from it.

References 1. Aggarwal, C.C. and Wang, H., Text Mining in Social Networks, in: Social Network Data Analytics, pp. 353– 378, 2011. 2. Balahur, A., Sentiment Analysis in Social Media Texts. Proceedings of the 4th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis, pp. 120–128. 3. Berry, M.W. (Ed.), Survey of Text Mining, vol. 2013, p. 2004, Springer-Verlag, New York, 2013. 4. Blanken, T.F., Van Der Zweerde, T., Van Straten, A., Van Someren, E.J.W., Borsboom, D., Lancee, J., Introducing Network Intervention Analysis to Investigate Sequential, SymptomSpecific Treatment Effects: A Demonstration in Co-Occurring Insomnia and Depression, in: Psychotherapy and Psychosomatics, pp. 1–3, 2019. 5. Miller, B.J., Veblen-Mortenson, J., Kunin-Batson, S., Nancy, A., Sherwood, E., French, S.A., A Bidirectional Analysis of Feeding Practices and Eating Behaviors in Parent/Child Dyads from Low-Income and Minority Households. J. Pediatr., 221, 93–98.e20, 2020. 6. Choudhury, D.M., Counts, S., Horvitz, E., Social media as a measurement tool of depression in populations. Proceedings of the 5th Annual ACM Web Science Conference, pp. 47–56, 2013, https://sci- hub.si/https://doi.org/10.1145/2464464.2464480. 7. Elhai, J.D., Hall, B.J., Erwin, M.C., Emotion regulation’s relationships with depression, anxiety, and stress due to imagined smartphone and social media loss. Psychiatry Res., 261, 28–34, 2018. 8. Younis, E.M.G., Sentiment Analysis and Text Mining for Social Media Microblogs using Open Source Tools: An Empirical Study. Int. J. Comput. Appl., 112, 5, (0975 – 8887), 2015. 9. Golden, J., Conroy, R.M., Bruce, I., Denihan, A., Greene, E., Kirby, M., Lawlor, B.A., Loneliness, social support networks, mood and wellbeing in community-dwelling elderly. Int. J. Geriatric Psychiatry, 24, 7, 694–700, 2009. 10. Hassan, A.U., Hussain, J., Hussain, M., Sadiq, M., Lee, S., Sentiment analysis of social networking sites (SNS) data using a machine learning approach for the measurement of depression. 2017 International Conference on Information and Communication Technology Convergence (ICTC), 2017. 11. Hasan, M., Rundensteiner, E., Agu, E., EMOTEX: Detecting Emotions in Twitter Messages, in: EMOTEX: Detecting Emotions in Twitter Messages, 2014. 12. He, W., Zha, S., Li, L., Social media competitive analysis and text mining: A case study in the pizza industry. Int. J. Inf. Manage., 33, 3, 464– 472, 2013. 13. Hernandez-Lallement, J., Gómez-Sotres, P., Carrillo, M., Towards a unified theory of emotional contagion in rodents—A meta-analysis, Neurosci. Biobehav. Rev., 2020, https://doi. org/10.1016/j.neubiorev.2020.09.010. 14. Jain, V.K., Kumar, S., Fernandes, S.L., Extraction of emotions from multilingual text using intelligent text processing and computational linguistics. J. Comput. Sci., 21, 316–326, 2017. 15. Keles, B., McCrae, N., Grealish, A., A systematic review: the influence of social media on depression, anxiety and psychological distress in adolescents. Int. J. Adolescence Youth, 25, 1, 79–93, 2020. 16. Kupferberg, A., Bicks, L., Hasler, G., Social functioning in major depressive disorder. Neurosci. Biobehav. Rev., 69, 313–332, 2016.

Extraction of Depression Symptoms  321 17. Ma, L., Wang, Z., Zhang, Y., Extracting Depression Symptoms from Social Networks and Web Blogs via Text Mining, in: Part of the Lecture Notes in Computer Science book series, vol. 10330, Springer, Cham, LNCS, 2017. 18. Muhammad, A., Wiratunga, N., Lothian, R., Contextual sentiment analysis for social media genres, in: Knowledge-Based Systems, vol. 108, pp. 92–101, 2016. 19. Neri, F., Aliprandi, C., Capeci, F., Cuadros, M., By, T., Sentiment Analysis on Social Media. 2012 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, 2012. 20. Nguyen, T.H., Shirai, K., Velcin, J., Sentiment analysis on social media for stock movement prediction. Expert Syst. Appl., 42, 24, 9603–9611, 2015. 21. Nguyen, T., Phung, D., Dao, B., Venkatesh, S., Berk, M., Affective and Content Analysis of Online Depression Communities. IEEE Trans. Affective Comput., 5, 3, 217–226, 2014. 22. Palhano-Fontes, F., Barreto, D., Onias, H., Andrade, K.C., Novaes, M.M., Pessoa, J.A., Araújo, D.B., Rapid antidepressant effects of the psychedelic ayahuasca in treatment-resistant depression: a randomized placebo-controlled trial. Psychol. Med., 49, 4, 655–663, 2019. 23. Park, M., McDonald, W.D., Cha, M., Perception Differences between the Depressed and NonDepressed Users on Twitter. Proceedings of the Seventh International AAAI Conference on Weblogs and Social Media, 2013. 24. Peng, Z., Hu, Q., Dang, J., Multi-kernel SVM based depression recognition using social media data. Int. J. Mach. Learn. Cyber., 10, 43–57, 2019, https://doi.org/10.1007/s13042-017-0697-1. 25. Salloum, S.A., Emran, M.A., Monem, A.A., Shaalan, K.A., Survey of Text Mining in Social Media: Facebook and Twitter Perspectives. Advances in Science. Technol. Eng. Syst. J., 2, 1, 127–133, 2017. 26. Spurk, D., Hirschi, A., Wang, M., Valero, D., Kauffeld, S., Latent profile analysis: A review and “how to” guide of its application within vocational behavior research. J. Vocational Behav., 120, 2020, https://doi.org/10.1016/j.jvb.2020.103445. 27. Wang, X., Zhang, C., Ji, Y., Sun, L., Wu, L., Bao, Z., A Depression Detection Model Based on Sentiment Analysis in Micro-blog Social Network, in: Trends and Applications in Knowledge Discovery and Data Mining, pp. 201–213, 2013. 28. Zucco, C., Calabrese, B., Cannataro, M., Sentiment analysis and affective computing for depression monitoring. 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2017.

Part 3 CYBERSECURITY

18 Fog Computing Perspective: Technical Trends, Security Practices, and Recommendations C. Kaviyazhiny*, P. Shanthi Bala and A.S. Gowri Department of Computer Science, School of Engineering and Technology, Pondicherry University, Puducherry, India

Abstract

As cloud computing is not viable for many Internet of Things (IoT) applications, fog computing is an emerging technology that inherits cloud computing platform and addresses the need for IoT and industry IoT. It reduces the time taken for communication between IoT and the cloud and substantially reduces the bandwidth that affects the IoT performance. The incorporation of IoT applications in fog paves way for a high chance of vulnerabilities in fog. So, the fog layer requires more security to protect the data both in transit and rest. Hence, the security issues of fog computing and the existing solutions are revealed in this chapter. It has been concluded that fog computing provides better performance than its counterparts like Edge Computing, Cloudlet, and Micro-Datacenter. This chapter investigates major security issues in fog computing and provides possible solutions and security recommendations to meet IoT security goals. Keywords:  Fog computing, threats, Internet of Things, machine learning, CISCO IOx, reference architecture, security solution

18.1 Introduction Fog is a computing paradigm that resides between the edge and cloud environment and provides storage and computation services in a distributed way. Fog computing is capable of connecting heterogeneous devices that process data on the edge of the network. Fog computing is also termed as OpenFog Consortium [1]. Fog computing is a decentralized environment and supports mobility. It helps to improve the performance of cloud computing infrastructure by minimizing the requirements to process and store data. Fog computing plays a vital role in the Internet of Things (IoT) and overcomes the issues by connecting multiple edge nodes directly to the physical devices [2]. Fog computing is highly prompted to support resource-intensive IoT applications that require low latency. It also requires computing devices which store a large volume of data and effectively process the acquired data.

*Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (325–352) © 2021 Scrivener Publishing LLC

325

326  The Smart Cyber Ecosystem for Sustainable Development IoT device is the one that includes both hardware and software. For sensing and monitoring purposes, it uses sensors and actuators. The appropriate software is integrated with the hardware for smooth functioning. The devices are operated through the Internet and enable the data transfer between devices automatically without human intervention. Thus, it requires fast decision-making techniques to settle down scalability and reliability issues. The fog model provides a scalable decentralized solution for all the issues that are faced in the IoT [3]. The fog paradigm is hierarchically distributed and provides a local platform between the edge devices, i.e., IoT devices and the cloud environment. It has the provision to filter, cluster, manipulate, analyze the transmitted data, and thus provide instant responsiveness. The growth of IoT leads to large volume of data generation. The IoT applications like healthcare and smart cities generate 30 million patient data at the rate of 25,000 tuples per second and millions of tuples per second [8], respectively. In 2019, the data generated by machines and humans may reach 500 zettabytes. CISCO submitted a report that in 2020 around 50 billion devices would connect to the Internet, and by 2025, it would be 500 billion [5, 10]. This prediction demands a fog environment to improve its optimization and service provisions. It is also necessary to balance security and privacy issues [4]. The huge volume of IoT data leads to heavy traffic congestion and difficulty in processing. Early IoT framework is directly connected with the cloud platform, and it has some drawback which cannot solve all the problems. IoT applications need mobility support, location identification, and geo-distribution, and those features cannot be achieved by the cloud layer alone; this arises a new computing paradigm called fog which keeps the data and resources near to the edge of the network that leads to the new development of applications [5, 6]. Fog is similar to cloud computing in terms of features and characteristics. Fog computing can be deployed in various fog service providers that could not be trusted fully and has many vulnerabilities. Some challenges in fog can be solved by using existing methodology, but still, it has some issues due to its unique characteristics such as heterogeneous node, mobility, massive data storage, and latency [13]. The proposed chapter discusses various issues, the need for cross industrialization of fog, the technological trends that support the fog environment, and various recommendations provided to enhance the security of the fog paradigm. Fog computing can be integrated with IoT which creates new trends in the services called Fog as a Services (FaaS). Fog as a service produces a series of fog node which can be located amidst the IoT layer and the cloud layer. The series of fog nodes are connected with each other and able to perform local storage, computing, and networking (elastic resources) [14, 20]. FaaS allows to develop a new model for the business and service to the customer. Cloud computing may suitable for big companies only as it uses private cloud and it has to store and process huge volume of data but FaaS is suitable for both large as well as small companies to deploy public or private fog services and compute the small volume of data to satisfy the needs of all customers [15].

18.2 Characteristics of Fog Computing The fog paradigm is a hierarchical model that distributes the IoT application at the edge of the network by distributing all the edge resources. The foremost feature of fog computing is IoT data processing, it processes all IoT data locally by keeping all the fog nodes nearby the

Fog Computing Perspective  327 user module. While comparing cloud computing with fog computing, it has some distinct features which make fog more efficient [16, 17]. Figure 18.1 depicts the various characteristics of fog computing. Location Awareness: Fog nodes are deployed in a diverse location, and these nodes can trace actively or passively to provide the best services to users. As the end devices, i.e., IoT devices, are nearby the fog node, it can able to identify the location of the end devices based on the fog node location. Low Latency: Fog node will provide services and take decisions by analyzing the local data that are fetched from the end devices rather than accessing and deciding in the cloud layer and also fog nodes are proximate to the end devices. Hence, it provides low latency compared to the cloud. Support Large Scale IoT Application: Fog will support large scale industrial applications. Due to the centralized structure of industries, it does not provide power grid management, environment management, water, and weather monitoring but fog computing is autonomous and independent to manage all the industrial devices. Predominance Wireless Access: The connection between the IoT, fog nodes, and cloud layer is wireless and it should provide predominance access to all the hierarchical layers and the access in all the layers should also be secure. Decentralization: One of the uniqueness of fog is decentralization. No centralized servers are managed in the fog network. As the fog node is distributed in different geographical areas, centralized architecture does not provide many services. Due to this decentralized behavior of fog node, it can easily cope with provisioning resources to all the fog nodes in the large network. Geographical Distribution: Fog nodes should be deployed in many locations because these nodes are the intermediate nodes to all the end devices and the cloud layer. There

Location awareness Low latency

Heterogeneity

Mobility

Characteristic of Fog computing

Predominance wireless access

Geographical distribution Decentralized

Figure 18.1  Characteristics of fog computing.

IoT application

328  The Smart Cyber Ecosystem for Sustainable Development should be an equal distribution of bandwidth to the entire layer. All the layers should receive a high-quality data stream even the devices have more number of nodes in between. Mobility: In many applications, fog nodes need to connect with the mobile devices directly and its architecture is decentralized in nature so it provides better mobility services. It uses some mobility protocols like location ID separation protocol (LISP) to obtain mobility services. Heterogeneity: End devices and fog nodes are manufactured by different vendors and the data gathered from the end devices are heterogeneous. So, the fog node should capable of process and manipulate these heterogeneous data and it should also work on different platforms.

18.3 Reference Architecture of Fog Computing The reference architecture of fog computing contains four layers. Edge devices, gateway, and IoT like sensors and actuators are present in the bottom layer. Some apps and software are installed in the end devices to enhance the features. The network and fog nodes are in the next layer; through this layer, the end devices can communicate with the cloud layer [7]. It acts as an intermediate layer in the architecture. The next is the cloud layer which contains all the cloud services, resource management, and support the processing. The resource management layer is the top most layer that helps to optimize the usage of cloud and fog services. The main goal of this layer is to minimize the cost of using the cloud and also to reduce latency by distributing or pushing some tasks to the fog nodes [23]. This layer also holds the quality of services for fog computing that leads to ingenious and smart applications for users. Figure 18.2 describes the reference architecture of fog computing. The resource management layer contains the following services like data flow and task scheduling, knowledge base, raw data management, performance analysis, monitoring, resource allocation, profiling, and security [24]. Cloud services

Task Scheduling knowledge base Raw data management monitoring performance analysis

Resource

Resource management

allocation Profiling Security Network and gateway

Sensors and Apps

Figure 18.2  Reference architecture of fog computing.

Fog Computing Perspective  329 Flow and Task Scheduling: It will keep track of all the available resources and schedule the tasks based upon its flow and its executions. It interacts with resource provision service to identify the current number of tasks and flow in execution and predict the future allocation of resources and tasks. Knowledge Base: Knowledge base stores the historical data about the resource demand and application demands which helps to take a better decision-making for all the services. Raw Data Management: It has direct contact with the data source and with the help of the querying language (SQL and NoSQL), and it will access the raw data. MapReduce concepts are also used to get some relative data. Performance Analysis: The information from the knowledge base service is utilized, based upon the data it estimates the available resources in the cloud. The estimated information is used by resource provisioning services for its management, control of tasks, and data flow. Monitoring: It keeps track of all the application and services status for better performance analysis and sends it to other services during the requirements. Resource Allocation: It allocates network, fog, and cloud resources for the application. The allocation of resources is dynamic and get changed over time. It takes the decision based upon the information from the knowledge-based service and resource provisioning service. Profiling: It builds the profiling document based on the data collected from the various services (performance analyzing, knowledge base, and monitoring services). Security: It provides all the security measures like authentication, authorization, encryption, and decryption based on its need and requirement of the services and applications. The reference architecture converges and summarizes the overall operation and the functionality of the fog platform. It also discusses various storage management, resource management, and allocation techniques. The effectiveness of fog reference architecture is elaborated [7].

18.4 CISCO IOx Framework CISCO first introduced the term fog computing in many aspects to improve the scalability, reliability, and cost-effectiveness and proposed a framework to satisfy all the requirements is CISCO IOx. It provides all the services like a cloud computing platform. CISCO brings the cloud features near the edge node through its frameworks. CISCO IOx (Internetwork operating system) provides consistency and uniform capabilities for application across the network infrastructure. It allows to execute IoT applications within the fog layer. Fog computing comprises of three services: Infrastructure-as-a-Service (IaaS), Platform-as-a-Service (PaaS), and Software-as-a-Service (SaaS) [7]. The IaaS is created using CISCO IOx API, which contains Linux and CISCO IOS. Protocol limitations are eliminated through this CISCO IOx API usage. When an IoT application uses CISCO IOx API in the application layer, it means the fog layer can use any protocol to connect with the end devices without any restriction. PaaS is provided by Cisco DSX. Cisco DSX offers the following three features. Figure 18.3 describes the CISCO IOx architecture. 1. IoT environment consists of heterogeneous devices and the application should communicate with all the sensors and devices. Creating a different

330  The Smart Cyber Ecosystem for Sustainable Development Saas Application management & MaaS Paas CISCO DSx

Iaas CISCO IOx + Linux

Figure 18.3  CISCO IOx architecture.

application based upon the sensor configuration support is impossible. So, it offers an abstract application for all kinds of devices. 2. It allows for developing a new environment with multiple language support. 3. Managing a large number of fog applications is difficult, so Cisco DSX makes this management task much easier. Cisco DSX also allows creating an interface between software-as-a-service and IoT devices. SaaS provides Maas (Machine as a Service) that allows defining functions and access to the users [26].

18.5 Security Practices in CISCO IOx CISCO IOx concentrates on security requirements like integrity, confidentiality, and authorization using the following CISCO IOx components [6]. 1. For the authentication, it uses fog director. The fog director checks the access control and allows the authorized user. It validates the access with the help of Enterprise Radius servers. 2. To check integrity CISCO IOx uses various mechanisms like application signature verification through (PEM format), OAuth API, Pluggable Authentication Module (PAM), and Trust Anchor Module (TAM). 3. For data confidentiality and privacy, it offers secure storage services (SSS) for the secure storage of keys, certificates, and data.

18.5.1 Potential Attacks on IoT Architecture Usually, the standard architecture consists of three-layer but in this section, the functionality and feature of each layer are further divided into one more layer for clear understandability. Each layer is analyzed separately and examine with its features and security issues. Authentication and access control are the major security aspects in all the layers. Each layer possess different attack based on their functionalities [3].

Fog Computing Perspective  331

18.5.2 Perception Layer (Sensing) The perpetual layer consists of the perpetual nodes or the sensors that are dynamically positioned. The perception layer is the lowest in the IoT architecture. This layer collects useful information or data from the environment like temperature, pressure, and humidity. Then, the collected data is transferred to the digital setup. In the perception layer, it uses the object for the unique identification and the low range of communication using the protocol RFID, Bluetooth, Near-field communication, and LoWPAN. The perception layer is also called a recognition layer. Power depletion attacks, physical tampering, jamming, signal interference and interception, firmware attack, and device authentication attack are some of the attacks in the perception layer. The ad hoc nature of this layer leads to various security threats that need to be addressed. Some important attacks that are frequently occurred in this layer are as follows. • Power Depletion Attack: The sensors are equipped with low power, which could drain quickly if there is an unauthorized intrusion; this vulnerability of draining the power is a power depletion attack. The battery power can also be rapidly drained by causing specific failure on the embedded circuits in the sensors. Many Denial-of-Service (DOS) attacks like the carousel attack which introduce routing loops and the stretch attack which makes long packets to drain the power. The Denial-of-Sleep attack makes the sensor node to drain power by keeping it awake. • Physical Tampering: The direct access to the chip by connecting to the signal wires or directly tampering with the chip itself remains a problem. This may cause a threat to the sensitive data in the chip to be accessed. It may also lead to damages or theft of the property. • Jamming: The jamming attack involves blocking the channel between the sender and the receiver or corrupting the message content with unwanted radio frequencies. This kind of DoS attack stops data from being communicated and it is a big constraint in wireless sensor networks. • Firmware Attack: The firmware is software used to rewrite the chip. By attacking the firmware, the chip gets corrupted. Some firmware attacks involve changing the firmware such that it cannot be used again. In the wireless network, the firmware can be attacked remotely. • Device Authentication Attack: Authentication attacks involve getting access to the device without proper credentials. They can be used to exploit the device or gain access to the data. A huge amount of authentication is handled in wireless sensor networks; a lightweight cryptographic algorithm is used to handle device authentication.

18.5.3 Network Layer The network layer is the core of the IoT architecture. It transmits data from the perception to the application layer. It may transmit it to other fog nodes, cloud, or directly to the other IoT devices itself. It provides secure data communication. Technologies used in this layer are mobile networks, ad hoc networks, and satellite networks. The network layer possesses certain network attacks that as follows.

332  The Smart Cyber Ecosystem for Sustainable Development • RPL Attack: RPL (Routing Protocol for Low-Power and Lossy Networks) occurs in distance vector routing of wireless sensor networks. The attack affects different functionalities of data transmission. The attacks include flooding, overloading of the routing table, rank attack increment, inconsistent DAG, and modification of version number. • Smurf Attack: The smurf attack is a Distributed Denial-of-Service (DDOS) attack done by generating numerous small ICMP packets with a targeted spoofed IP address. This attack often creates network traffic resulted from spoofed ICMP packet flooding and stalls the system of the victim. • Clone ID Attack: The ID of a node is replicated along with the cryptographic materials in the ID. The cloned node is then injected into the network. The cloned node forms the entry point to launch attacks and eavesdrop on the original node. • Sybil Attack: The Sybil attack is performed by creating many nodes with false identities in a network. It disturbs the routing mechanism of the network by disrupting the middleware decision by rigging votes in the group-based decision. • Sinkhole Attack: This attack is performed in the RPL network by modifying the rank value using a malicious node, which creates a lot of traffic. This makes the malicious node the preferred node often rather than optimizing the route. This affects network performance. • Black-Hole Attack: The black-hole attack is a variant of the DoS attack, where the malicious node attracts a lot of network traffic and then drops all the packets routed through the node. It is similar to the strategy of the Sinkhole attack, rather than degrading the network performance, and this attack eliminates all the packets. • Wormhole Attack: The wormhole attack is performed to affect routing patterns and degrade network quality. The attacker creates two malicious nodes that interchange the data received and eliminating the actual distant nodes to which the data needs to be routed.

18.5.4 Service Layer (Support) It provides a service platform for IoT devices. It has high computation and storage power for the devices. • Man-In-Middle Attack: The attack authenticates between two nodes and impersonates each node and establish itself as a genuine other node. This attack can lead to active eavesdropping and modification of data communicated. • Denial-of-Service Attack: The attack prevents the nodes from utilizing the service or resource in a network. • Flooding: It provided a huge amount of incomplete connection request packets that lead to denial of other packets requested to the service. This form of DOS attack is common and utilizes the memory and time of the service making it useless to other genuine requests.

Fog Computing Perspective  333

18.5.5 Application Layer (Interface) The application layer is the top most in IoT architecture. The objective of the layer is to reduce the between the user and IoT applications. The user can access all the resources through the application layer. It provides a user interface for the IoT applications. • Distributed Denial-of-Service Attack: DDOS attack occurs due to the presence of multiple DOS attacks which can alter the traffic of the whole network or a targeted resource. • Denial-of-Service Attack: DOS refers to a type of attack that prevents the nodes from utilizing the service or resource in a network.

18.6 Security Issues in Fog Computing Fog is inheriting the characteristics and features of cloud computing. Hence, fog computing also has the same security issues as cloud computing. Cloud computing has more vulnerability in terms of security because it is used to compute and store the data in a centralized server and it provides the way to the hackers for the security attacks. The important factor is that it has a concern with cloud computing security measures. Comparing with cloud computing, fog computing has some enhanced security measures because of its key features [32]. The main feature is that the fog node is located near the data sources so it minimizes the internet connection dependency. Local data processing and analysis in the fog node reduces the security attack during the data transmission. Secondly, it increases the data security in the fog node itself; hence, it eliminates data theft through the network. Though fog computing has so many security enhancement features, it inherits the cloud computing characteristic and paradigm; so still, it has some limitations and security attacks. The following section covers the possible attacks/threats and solutions for those issues.

18.6.1 Virtualization Issues Virtualization allows creating replica or likeness of the hardware, software, and network virtually. It uses a virtual machine monitor to monitor the hardware as well as the software separately. A virtual machine monitor is also called a hypervisor. The hypervisor creates a virtual machine and terminate and allocate the resources. It makes computing tasks much easier and flexible. The role of virtualization is important in the fog computing paradigm. Though virtualization provides various advantages, it creates some challenges and threats in fog computing. Some of them are as follows. Hypervisor Attacks: In hypervisor attacks, a virtual machine monitor allows the attacker to hijack the virtual machine. It increasingly allocates resources to the virtual machine which leads to some denial of service and deadlock on the virtual host. VM-Based Attacks: The intruder might use a malicious virtual machine which may control and access other virtual machines in the virtual host. It may get access to sensitive data and make it unprotected. It will collapse the operation system and modify its configuration which results in shut down the system forcefully and improper communication between virtual machines.

334  The Smart Cyber Ecosystem for Sustainable Development Weak or No Logical Segregation: It helps to partition the network into a small network through which it develops great control and communication between virtual hosts. The objective is to provide authorized data access to the network and it restricts the data security over this non-logical segregated network. Side-Channel Attack: This is the hidden channel in the virtual environment which stores the data cache for the future data recovery purpose. The attackers exploit data access. Privilege Escalation Attacks: The system may have some errors/bugs in programs and design flaws, the attacker gains an advantage over these and will have malicious access to the networks and its applications. Possible Solution/Technique for the Virtualization Issues: • Multifactor authentication: It provides more than one authentication mechanism to identify the authorized user. • Intrusion Detection System: It monitors the network traffic, detects the malicious activity, and intimates/alerts before it produces any damage. • User data isolation: It hides and segregates the user data(one user cannot access another’s one). • Role-Based Access Control model: It will limit the access to the network-based on the user role. • User-based permissions model: It will restrict software usage by unauthorized users. • Process isolation: It isolates the process procedure and execution flow by a certain group of users.

18.6.2 Web Security Issues The web is an important component of any organization; most of the confidential tasks are transferred and viewed through the web. If the web or a server weak, then the sensitive data may get leaked. So, security is a primary component of a web. The proper designing of codes for the web and the data being protected will ensure the safety of data. Web security is also known as cybersecurity that protects against unauthorized access to data on the web. The threats and attacks under web security are as follows. SQL Injection: It injects or adds a query to the database server behind the web. The injected query will allow the hacker to access the database. They can modify, read, and even delete any important record in the database. It is difficult to identify the malicious query in the database after this attack is done. Cross-Site Scripting: It adds a malicious script on the client-side which may redirect the web page. The attacks can access website cookies, session tokens, and sensitive data. Crosssite scripting happens in front-end scripting. Cross-Site Request Forgery: It forces the end-user to transmit unaccredited command; because of this unauthorized command, clients may lose their trust. The command will induce the website to do money transactions, sensitive data transmission, and access cookies on the website. This attack is also called a one-click attack/session riding.

Fog Computing Perspective  335 Insecure Direct Object References: The user will give a valid/authorized input to the web as the result of the program; it provides the user to access the object directly. This provides vulnerability to the website. The hackers also follow the same strategy to get direct access to the object. They may also provide valid input (like database files and records) to gain object accessibility and modify the object parameters, features, and operation. Malicious Redirections: It inserts the code on the website and it will redirect the website based on the attacker’s need. The redirection of the website will provide some unwanted advertisement displaying and also leads to some dangerous impact on the websites by downloading and installing some malicious API. Drive-By Attacks: It triggers and redirects the website to install some dangerous API and it will run in the background without the user’s knowledge. The user cannot know the background running application which may lead to account/session hijacking. Possible Solution/Technique for Web Security Issues: • Secure code: The code has to adopt a secure coding standard. It should automatically deny the code access from the attackers. It should provide a heed compiler warning. • Find and patch vulnerabilities: The code may contain a patch that provides vulnerability to the website. It should be detected and prevented before any attacks. • Regular software updates: The software should be updated regularly to get the recently updated security enhancement feature in it. • Periodic auditing: It will keep track of all the activities that are happening on websites like amount transaction, updates, installation, etc. With the help of auditing, it will identify the possible threats that may happen in the future. • Firewall: Proper usage of firewalls is essential. It will block unnecessary website access. Through this, it will eliminate system redirection because of the malicious attacks. • Anti-virus protection: Best anti-virus protection tools should be installed in the system to prevent the unwanted installation of software. • Intrusion Prevention System: It will monitor and keep track of the website to avoid unwanted activity happening on the web. It suspects some activity, tracks its behavior, blocks it, and finally reports it to the user.

18.6.3 Internal/External Communication Issues Data transmission and exchange plays a major role in the organization. Internal communication takes within the organization and external communication happens to the environment in which the organization operates. In internal and external communication, threats may arise through network security threats, human threats, and software threats. Man-in-the-Middle Attack: It is an active attack where the malicious intruder acts between the sender and the receivers. The intermediate intruder has direct contact with both the parties so they may assign and give illegal commands.

336  The Smart Cyber Ecosystem for Sustainable Development Single Point of Failure: Failure of a particular part of the system halts the entire system from working. The entire system will collapse due to a single point of failure. It will destroy the internal and external communication between the organizations. Possible Solution/Technique for the Internal/External Communication Issues • Encrypted communication: Data should be encrypted before they are communicated or transmitted in the network. • Mutual/multi-factor authentication: It provides more than one authentication mechanism to identify the authenticated user using security tokens, biometrics, one-time passwords. • Partial encryption: It will encrypt only the part/certain portion of the original data then the remaining part remains unchanged. • Isolating compromised nodes: The nodes which are detected or identified as a malicious node are isolated or removed from the network. • Certificate pinning: It is a technique to identify the authenticated system or server in the network. The certificate pinning is used to evade a man-in-themiddle attack on the client-side even after an SSL handshake. • Limiting the number of connections: Only a minimal number of connections should be established in the communication process. • Transport layer security (TLS): It is a protocol that provides a security mechanism over the client-server architecture. It offers integrity and privacy during data transmission.

18.6.4 Data Security Related Issues Data security should provide authenticated access to data from a database or system. Even though many data security measures are used in the computing process, many threats and attacks are arising in day-to-day life. Lack of data security will corrupt the entire data. Data security must be taken care of in every phase like storage, transferring data in the network, and providing access credentials. Data Replication and Sharing: It is a passive attack; it will not alter the original data but the attacker will duplicate data for malicious activity. The user may not know data replication and sharing, only they will come to know after the issues arise. Data Altering and Erasing Attacks: Here, the hacker will modify the data or delete it before it is transmitted. Due to this modification, it produces misinterpretation in the organization. Possible Solution/Technique for Data Security Issues: • Policy enforcement: Proper policy enforcement technique is maintained. It should hold the access credential, predefined policies, and protocols used in the data security.

Fog Computing Perspective  337 • Security inside design architecture: It addresses all the potential risks in the environment and techniques to be followed to avoid security risk in the future are also deployed in the design. • Encryption: Data is converted into ciphertext so that the attackers cannot gain knowledge about the data even though they access the data. A proper encryption technique is used. • Secure key management: Encryption and data transmission process require key (public and private). The key should also be secured using a standard encryption technique. • Obfuscation: This should make the data unclear to the hackers. • Data Masking: Most sensitive data should be masked or hide based on its requirement.

18.6.5 Wireless Security Issues Most of the communication takes place using wireless/connectionless. The data is transmitted through the network to reach its destination. Before reaching the destination location many network security threats and issues may arise. Network-based issues are also associated with wireless security issues when the data is transmitted over the network. The main and menacing attacks are as follows. Message Replay Attack: It is the network attack where the attackers receive the data and replay it. For example, the sender sends the message, instead of the intended receiver the attacker gets the message and replays to the sender that the message reaches the destination. Message Distortion Issues: It is a misinterpretation by the receiver; hence, the receiver carries out the process incorrectly. Here, the sender and the receiver’s thoughts are unbalanced. Data Loss: The data is theft or corrupted or even deleted by the attacker due to the data transmission. This makes the software or application to unreadable the data. Data Breach: It happens intentionally or unintentionally but it will release the data to the untrusted/malicious environment which may produce menacing effects. Sniffing Attacks: It is a kind of network attack which predicts and analyzes the network traffic and then captures the data during the transmission through the network. Possible Solution/Technique For Wireless Security Issues: • Authentication: It will verify the authenticity of the user. Many authentication mechanisms are available like mutual authentication, biometric, image authentication, and token-based authentication. • Encrypted communication: Data is converted into ciphertext so that the hackers cannot gain knowledge about the data even though they access the data. A proper encryption technique is used. • Key management service: Encryption and data transmission process require key (public and private). The key should also be secured using a standard encryption technique. • Secure routing: Proper routing protocols are used like RIP, EIGRP, and OSPF

338  The Smart Cyber Ecosystem for Sustainable Development • Private network: Private IP addresses and the virtual network are used for transmission so it will increase the security mechanism. • Wireless security protocols should be used.

18.6.6 Malware Protection It is the most necessary thing in the devices. Malware is installed and perform many unnecessary processes without user knowledge. Malware protection is used to prevent deleterious and harmful attacks. Some of the malware are as follows. Virus: It is a deleterious program that leads to replicate the data, copies it, and makes the system to halt. The main motive of a virus is to affect the vulnerable system and get access to sensitive data. Mostly, virus spreads through mail attachments. Trojans: Trojan is a malicious code or software that controls the system and even damage it. It uses the Trojan file to separate the malicious action. Worms: The worm is the malware that replicates rapidly and affects other systems in the network. It uses a network to travel from one system to another. It does not infect the file. Ransomware: It is the type of malware that steals user data and accessibility. It demands the user to pay the amount to unblock the action. It transfers through mails, web-based messages, and attachments. Spyware: Spyware monitor the system activity. It requires an internet connection and it is difficult to identify even by anti-virus software. Possible Solution/Technique For Malware Protection: • Anti-malware programs: It is the software that safeguards the system from various malicious attacks. It will detect the attack and eliminate it automatically. • Intrusion Detection System: It monitors the network traffic, detects the malicious activity, and intimates/alerts before it produces any damage. • Rigorous data backups: The prevention technique which can able to rescue the system even after the malware attack. It is like a recovery store and it can be used after the data loss.

18.7 Machine Learning for Secure Fog Computing The machine learning algorithm is used to improve the fog node security based on its analysis. Data analysis can be done in two ways one is a reactive analysis and another one is predictive analysis [33]. The reactive analysis is always performed nearby the edge layer, i.e., on the sensor layer. The predictive analysis process is performed in the top most layer (cloud layer). Predictive analysis requires more computational capability. The premise is that as the cloud has more computational power, it is preferable to use predictive analysis in the cloud. The fog computing follows a hierarchical structure. Layer 1 is the cloud layer and layer 4 is a sensor or IoT layer. The remaining intermediate two layers are fog layers. Layer 3 is fog nodes for the neighborhood, and it collects the raw data from the sensor layer. Layer 3 is responsible for two functionalities. Figure 18.4 describes the fog architecture.

Fog Computing Perspective  339 Cloud

Layer 2 Fog nodes for the community

Fog

Fog

Fog

Layer 1 Cloud

Fog

IoT environment

Fog

Layer 3 Fog nodes for the neighborhood

Layer 4 Sensor

Figure 18.4  Fog architecture.

1. It will identify the patterns of threat from the raw data using a machine learning algorithm. 2. It will perform feature extraction from the collected data and will minimize the amount of data that is transmitted to the upper layer. Layer 2 is the fog node for the community that is responsible for data collection from layer 3. This layer uses a Hidden Markov Model (HMM) and maximum a posterior (MAP) algorithms to predict the hazards activities.

18.7.1 Layer 1 Cloud It is the first layer in the secure fog computing paradigm, and it uses techniques like ANN, deep learning, ANN, reinforcement learning, decision trees, and Bayesian network algorithms for the security aspects. • ANN: Artificial neural network is the computational model, it has three layers. The first layer contains the input neuron. The intermediate layer is responsible for the computing function. The intermediate layer may have multiple hidden layers based on its computational purpose. The layer uses the random function to learn the features from the input. Rather than using the entire data sets, it only prefers to use the data samples that make ANN more advantageous. The third layer provides the output based on the data analysis made in the second layer. It simply learns the data sample and produces the output using the light mathematical model. • Deep Learning: Deep learning performs data analysis tasks on the unstructured data that makes the IoT environment more effective. It provides a high level of data abstraction using multiple non-linear transformations. Deep

340  The Smart Cyber Ecosystem for Sustainable Development learning is the subfield of ANN, and it provides better decision-making techniques to improve the security in the cloud layer. • Reinforcement Learning: Reinforcement learning is the dynamic programming that trains the system module by its reward and penalty. If the algorithm provides a correct output, then the user rewards the system based on that algorithm specification likewise for the wrong analyses the penalty is made to the system to reconfigure the work. This algorithm is based on the experimental practice that suits the real-time environment like fog platform and IoT. • Bayesian Networks: Bayesian network is a probabilistic graphical model that predicts the cause based on the contribution factor. It provides the probabilistic relationship between the cause and the event. This graph-based prediction technique is used in fog networking to prevent attacks. • Decision Tree: A decision tree is a supervised machine learning technique. It classifies the data based on the requirement and its parameters. It splits the data into a decision node and the leaf node. The leaf node is the final decision outcome of the data.

18.7.2 Layer 2 Fog Nodes For The Community Layer 2 uses the machine learning algorithms like MAP, regression, ANN, HMM, and decision trees for the predictive analysis process. • Hidden Markov Model (HMM): It is a statistical model that is used in the application of reinforcement learning and temporal pattern recognition. It is a kind of unsupervised learning. It uses the probabilistic approach to detect and analyze the probability of event occurrence. • Maximum A Posteriori (MAP): It is a graphical model that calculates mean and variance in the dataset and is used as a Gaussian parameter. It comes up with Bayesian settings and Bayes rule. • Regression: It is supervised learning. This model is used when the dataset is a continuous value. Mostly the data generated by the sensor nodes are continuous and temporal data, hence the regression model is used to find the linear correlation between the input (x-axis) dataset and the output (y-axis).

18.7.3 Layer 3 Fog Node for Their Neighborhood This layer fetches the data from layer 2 and uses a certain machine learning algorithm for the reactive analysis process. • KNN Algorithm: K-nearest neighbor classifies the dataset into separate groups based on their parameters and behavior. For the group segregation, it uses the distance formula to calculate its behavior. • Naïve Bayes: It uses the Bayes theorem to classify the dataset. Naive Bayes techniques often suit for the high dimensional inputs. Probability technique like prior probability is used. Prior probability uses the value from the

Fog Computing Perspective  341 previous probabilistic experience. It also analysis the historical classification prediction techniques. • Random Forest: It is a popular machine learning technique for complex data sets. It will generate a random number of decision trees based on the variable set in the data. It exactly predicts and isolates the knowledge from the data analysis process. • DBSCAN: It is a clustering technique used in the group of closer points with minimal distance. With DBSCAN, the association between the data points can be found. Predicting the hidden pattern and trends in the data will be easier. It will also remove noise from the data. For measuring the distance between the points Euclidean distance method is used. DBSCAN requires two-parameter eps (epsilon) and midpoints (minimum number of points). Based on these constraints, the clustering process takes place.

18.7.4 Layer 4 Sensors In layer 4, no machine learning algorithm is used. This layer is also called the IoT environment. Here sensor monitored and send the sensed data to layer 3 for the further process.

18.8 Existing Security Solution in Fog Computing The following section discusses the best existing security solution for fog computing. It explains all the techniques and principles used in the security solutions and where the solution is applicable to use is also suggested.

18.8.1 Privacy-Preserving in Fog Computing The fog layer resides between the end devices (sensors) and the cloud layer, so to secure the sensor data, the following steps have to be followed. Initially, the data is collected and the feature is extracted from feature extraction algorithm. Make the data fuzz by including the Gaussian noise to the data, and it will minimize the chance of security attacks. To avoid a man-in-the-middle attack, the data is partitioned into the block, and then, the block is shuffled and send through the network. For each block of data, separate public key infrastructure for encryption should be maintained [29]. While transmitting the data packets, the packets should be encrypted and reordered. On the receiver side, the partitioned block is rearranged and decrypted to ensure its security. To improve the above, privacy-preserving techniques should select the best encryption algorithm and key management algorithm. Before and after the data transmission process, data manipulation process (encryption, decryption, segregation, data partitioning, and huffing) should be maintained. Sensors are generating data continuously; so in addition to this, implementing the privacy-preserving mechanism may crash and overload the fog system. Figure 18.5 explains the privacy-preserving mechanism. AES algorithm and data partitioning is used for the data encryption and data partitioning mechanism [35, 36]. Pseudocode for the privacy preservation in fog is discussed in the following section.

342  The Smart Cyber Ecosystem for Sustainable Development Start

Fog node reads dataset

Apply PCA to extract required feature & assign the selcted features in feature

Add Gaussian noise to feature

Use AES algorithm to encrypt feature cvs into feature

Divide the encypted file into n-number of blocks by secret cipher share algorithm

Transmit the blocks to the cloud irrespective of the order

Received blocks are rearranged, decrypted on cloud layer

End

Figure 18.5  Flow chart for privacy preservation mechanism.

18.8.2 Pseudocode for Privacy Preserving in Fog Computing // It is assumed that dataset.csv is collected from IoT devices. //FN calls the procedures individually for feature extraction, adding Gaussian noise, Encryption and partitioning. Input: Fog node(FN) reads dataset.csv Output: The encrypted data sent to the cloud.

Fog Computing Perspective  343 Begin read dataset.csv feature_extraction() Gaussian_noise() AES_encryption() Block_partition() Upload the partitioned blocks b1.......bm to cloud transmission to cloud end. End

//

18.8.3 Pseudocode for Feature Extraction Feature_extraction() { Input: dataset.csv Output: feature.csv // this procedure selects only m_required attributes Begin A D-dimensional training set X={ x1,x2,… …xN} and the new (lower) dimensionality d (with d≤ D) N 1 xi Compute the mean x i=1 N N 1 Compute the covariance matrix Cov(x)= (xi -xi)(xi -xi)T i=1 N Find the spectral decomposition of Cov(x), obtaining the eigenvectors Є1, Є2,… … ЄD and their corresponding eigenvalues λ1, λ2, … .λD. Note that the eigenvalues are sorted, such that λ1 ≥ λ2 ≥ … ≥ λD ≥ 0 For any xЄ RD, its new lower-dimensional representation is:

Σ

_



Σ

_

_

Y = ЄT1(x − x),Є2T(x − x),..............,ЄdT (x − x))TЄRd, And the original x can be approximated as _

_

_

X ≈ x– +(Є1T(x − x))Є1 + Є2T(x − x))Є2+ ... +ЄdT (x − x))Єd return feature.csv end }

18.8.4 Pseudocode for Adding Gaussian Noise to the Extracted Feature Procedure for Gaussian_noise( ) { input: feature.csv output: attributes of feature.csv associated with Gaussian noise.

344  The Smart Cyber Ecosystem for Sustainable Development Begin Y= f(X)+ Є // f(X) is a feature extracted data from feature extraction algorithm Noise=xi+ Gaussian(0,σi2) // Є is the Gaussian noise Return(y) // y is the noise added feature set End }

18.8.5 Pseudocode for Encrypting Data Procedure for AES_encryption( ) { Input: y // the noise added feature set Output: Encrypted feature set(y) Begin Cipher(byte in[4*wb,byte out[4*wb], (round+1)*wb-1]) Begin Byte state[4,wb] State = in AddRoundkey(state,w[0, wb-1]) For round = 1 step 1 to wn-1 subBytes(state) shiftRows(state) Mixcolumns(state) AddRoundkey(state,w[round*wb, (round+1*wb-1]) end for subBytes(state) shiftRows(state) AddRoundKEY(State, w[wb*wn, (wn+1)*wb-1]) Out=state End }

18.8.6 Pseudocode for Data Partitioning Procedure for Block_partition() // the encrypted feature set is divided into m-blocks by secretcipher share algorithm Input: encrypted_y.enc Output: (b1, b2, b3, …..bm) blocks of encrypted_y.enc { Begin Pv΄ // secondary public vector attr // to be overshared attribute ks,ps // modulus and key of the oversharing server S Main: /* retrieving the current secondary public vector */

Fog Computing Perspective  345 pv˝= (SELECT PV2 FROM public_vectors WHERE AttributeName = attr) // current secondary distribution vector W˝= Dks (pv˝ mod ps) W΄= Dks (pv’attr mode ps) // new secondary distribution vector /* updating the secondary public vector*/ UPDATE public_vectors SET PV2= PV’attr WHERE AttributeName = attr for all tuples do v1 = share1(shares(attr)) v2 = share2 (shares(attr)) shares(v) = Reconstruct((v1, v2), w”)



X’ (share1(shares(v)),share2(shares(v))= SSshare (2,2) s(v)

end for end }

18.8.7 Encryption Algorithms in Fog Computing Poor authentication protocols between the IoT and the fog layer bring many security-related issues. Fog computing does not have the best authentication and secure communication protocols in the present days. The encryption techniques like fully Homomorphic and Fan-vercauteren are suggested for the data security [32]. Homomorphic encryption will encrypt the data and it will reduce the key distribution to improve security. Privacy-Preserving Aggregation (EPPA) scheme also provides a good encryption process with low authentication cost. Comparing to all the encryption algorithms, AES algorithm suits best for both the IoT device’s data as well as the fog systems. For both encryption and decryption processes, the AES algorithm uses less time and memory utilization. The size of the encryption key acts a vital role in the AES algorithm which gives additional advantage to the AES. For better security enhancement, selecting the appropriate algorithm that suits the environment and requirement is essential.

18.9 Recommendation and Future Enhancement Security is vital in all domains. To store the data securely, the storage process should follow some predefined security protocols. To prevent the system from various attacks, then it should follow security rules and policies. The following section provided the security recommendation that has to follow in advance to stay away from malicious issues and the future enhancement techniques to inflate the system efficiency.

18.9.1 Data Encryption Data encryption checks the confidentiality aspect of the data. In IoT, the device does not have much memory to store data locally, and the encryption process within the device itself is not advisable. Hence, fog computing overcomes all the disadvantages and provides an encrypted data transmission. Though fog nodes allow encrypting the data, encrypting all the collected data is erroneous. Only necessary or sensitive data should be encrypted and

346  The Smart Cyber Ecosystem for Sustainable Development send through the network. Several techniques have been used to validate the received data. For the IoT devices, the AES algorithm is efficient for the encryption process because of less CPU utilization. The secure socket layer (SSL) protocol provides secure communication between the client and server [26]. The fog system designer should access an ample amount of data and provide sufficient security in the necessary environment. Recommendation 1: Data generated from the IoT devices (origin devices) need to be secured before it reaches the destination. Data is encrypted before it passes through the network. So, it assures secure data transmission between IoT devices, fog network, and cloud platform. Future Enhancement 1: To implement any data security mechanism, it requires more computational power. Hence, balancing node operation and security technique is essential. The encrypted data, i.e., ciphertext, needs more disk space when compared with the plain text so it reduces the overall working mechanism in all the layers.

18.9.2 Preventing from Cache Attacks To enhance the performance and power efficacy of other systems in the network, the fog node uses caching techniques. Cache memory stores the data of recently and most frequently used data. A cache can even store the user’s personal information. This leads to data theft and side-channel attacks. Some techniques are used to overcome these security issues are NEW CACHE and STEALTHMEM [27]. In the new cache technique, it comes up with some mechanisms like partition lock cache and random permutation cache. Smart meter data medication is the log record about the data transmission in the network. To avoid smart meter data medication in the metering architecture, the fog node has to store or retain the data for a specific time before it is transmitted. Even though the new techniques for cache attack prevention is expensive, fog platform developer concentrates on the cache implementation due to its various superiority. Recommendation2: Fog computing maintains the cache management system which leads to software cache attacks. It discloses the cryptographic key which provides the direction to the leakage of sensitive data. Future Enhancement 2: Elimination of cache attack is very expensive; the solution for the cache attack can abolish only a certain type of cache attacks. Cache attacks may arise in both hardware as well as at the software level. Hence, it is advisable to implement the cache attack prevention techniques in both hardware and software.

18.9.3 Network Monitoring Every fog platform must implement a network monitoring mechanism. The network monitoring system will help to detect the malicious attack in the early stage itself. The main functionality of network monitoring is to scan and monitor the network dynamically or it can divide the large network into small partitions based on the pre-defined rules and policies and then it will monitor easily. For an effective way of network monitoring, the fog platform uses the CLOUDWATCHER tool which partition and scan the network dynamically [27]. The scanning process can be either static, dynamic, or hybrid. An efficient way of scanning is attained through firewalls, anti-virus cum intrusion detection, and prevention system. To detect the threat in the large heterogeneous network, ANNs and rule matching techniques are used. A virtual private network (VPN) is implemented to isolate the network from external attacks.

Fog Computing Perspective  347 Recommendation 3: IoT devices generate data and they are processed in the fog layer and then transmitted to the cloud. Hence, the fog layer is continuously dealing with the private data. It should be monitored and analyzed for the occurrence of anomalous activity in the network by automated enforcement of rules and policies from communication security. Future enhancement 3: Fog network is connected to many small IoT devices; even though the data generated from each device is minimal, the number of fog node in the network is more, and handling all the generated data is difficult. Hence, the network should be partitioned and investigate separately to increase its processing and memory capacity.

18.9.4 Malware Protection Fog computing platform does not have sufficient malware protection as it is engaged in the continuous allocation of network and computational resources. Many fog systems are deployed in smartphones and tablets so there is a high possibility for malware injection. The injected malware will steal the user’s data, access, modify it, and even destroy it. The best solution for malware prevention in fog platforms is to deploy the physical malware detection devices and a tool like BARECLOUD used to predict the malware attacks [28]. Also, machine learning techniques like support vector machines with benign software is employed to detect the malware in the earlier stage. Recommendation 4: The Fog layer should be trained and protect themselves against the existing and new threats and attacks. Future Enhancement 4: Threat detection should be contrived with the help of network monitoring mechanisms and physical malware protection.

18.9.5 Wireless Security IoT devices like cameras, sensors, and healthcare monitoring are connected through a fog platform and transmit the data to its nearby nodes. To make the wireless data transmission secure, data communication should be encrypted using the Wi-Fi security algorithm. The Wi-Fi access point is visible to all nearby devices; hence, the access point should be secure enough to avoid attacks [29]. There is a high chance of vulnerability occurs in the network when the environment has a weak access point. Security attacks like Sybil attack and flood attack attackers become a part of the network, it will use the network bandwidth and network interruption. In healthcare monitoring application, an insecure wireless access point leads to many critical issues as it deals with patient health data, small changes in the patient data leads to a huge problem. Hence, standard wireless protocols should be implemented in fog computing applications. Recommendation 5: The fog layer should minimize the internal and external wireless communication between the end devices and destination platform. Future enhancement 5: Fog network comprises sensor networks and IoT devices so the fog layer should be hidden and secured, but it is difficult to maintain security in fog networks.

18.9.6 Secured Vehicular Network Fog computing supports vehicular networks. The vehicular network has a high dynamic topology when compared to a mobile network because in a vehicular network each node is a vehicle and it is moving dynamically so the change of network traffic and frequency

348  The Smart Cyber Ecosystem for Sustainable Development is drastic. To achieve a secure vehicular network in the fog platform each vehicular node should check user authentication, data confidentiality, integrity, and dynamic resource allocation. It should also maintain anonymous key management. Like the feature of Stamp [34], each node should be independent and stateless in its operation and securely keep the user location. The deployment of security measures in the fog-based vehicular network will avoid the attacks at the initial stage. Recommendation 6: To implement a secure vehicle network the node should protect itself against internal and external security threats. Future Enhancement 6: It has high variable network density, depending upon the network traffic density will increase or decrease. In both cases, there is the possibility of network disconnection which provides more opportunity for security attacks. So, network density should be maintained moderately.

18.9.7 Secure Multi-Tenancy Fog computing is the virtual environment and supports multi-tenancy. It is used to provide a pool of resources with security and privacy. To enhance the security in multi-tenancy, multi-factor authentication/mutual authentication, and logical data segregation, proper resource allocation techniques have been used. Fog platform also provides secure and Resilient networking services to adjust network topology. The cloud platform provides the best multi-tenancy with fewer security measures but fog overcomes the issues. Recommendation 7: Fair resource allocation mechanism should be used to protect against confidentiality and integrity and the fog layer should provide constrained access control on the data and network. Future Enhancement 7: Maintenance is difficult when the number of end-user shares and accesses the resource from the fog layer.

18.9.8 Backup and Recovery System backup and recovery will overcome all the system failures and security attacks. It is essential to store sensitive data in another location and use it when required. The data backup leads to many security problems [32]. It is easy for the attackers to attack the replicated copy of data. It is essential to monitor and prevent data backup storage from the replication before the recovery process. Some of the techniques are available to improve the data consistency are Rake Technology, Cold and Hot Backup Services Replacement Strategy (CBSRS), and Shared Backup Router Resources (SBRR). Recommendation 8: Based on the requirement, the fog layer should have a data backup and retrieval module. Future Enhancement 8: Data recovery and backup process are very costly and also there is a need to focus on data selection, mapping, and specifying access

18.9.9 Security with Performance Security in the system is an important factor but providing unnecessary security to all the resources is not advisable. Fog network capable of transmitting data loads, resource

Fog Computing Perspective  349 allocation, and sensors signal, so providing security in the network is essential but for the fog, the network is difficult because fog network does not have much computational power and security measure The security techniques will minimize the system performance and for implementing security in IoT devices it requires additional memory space, so this might be the reason for providing security based on its performance. Recommendation 9: The security mechanism is inversely proportional to the system functionality. When the security mechanism increases, functionality decreases. Hence, there should be a balance between the degree of functionality and security. Future Enhancement 9: The best security techniques are necessary to be chosen and it should not compromise the functionality of the system too.

18.10 Conclusion In this chapter, the various concepts of fog computing and its architecture have been discussed. The chapter identifies the weakness and loopholes of fog computing that are exploited by the attackers. The security challenges in IoT cum fog architecture are identified, and techniques to solve the vulnerabilities are suggested. It provided the machine learning algorithms used in the fog layer. The important features of fog computing are highlighted. Security challenges and future recommendation to enhance security in fog computing has been summarized.

References 1. Puliafito, C., Mingozzi, E., Longo, F., Puliafito, A., Rana, O., Fog computing for the internet of things: A Survey. ACM Trans. Internet Technol. (TOIT), 19, 2, 18, 2019. 2. Yi, S., Qin, Z., Li, Q., Security and privacy issues of fog computing: A survey. International conference on wireless algorithms, systems, and applications, Springer, Cham, pp. 685–695, 2015, August. 3. Rauf, A., Shaikh, R.A., Shah, A., Security and privacy for IoT and fog computing paradigm. 2018 15th Learning and Technology Conference (L&T), IEEE, pp. 96–101, 2018, February. 4. Abbasi, B.Z. and Shah, M.A., Fog computing: Security issues, solutions, and robust practices. 2017 23rd International Conference on Automation and Computing (ICAC), IEEE, pp. 1–6, 2017, September. 5. Ni, J., Zhang, K., Lin, X., Shen, X.S., Securing fog computing for internet of things applications: Challenges and solutions. IEEE Commun. Surv. Tutorials, 20, 1, 601–628, 2017. 6. Thota, C., Sundarasekar, R., Manogaran, G., Varatharajan, R., Priyan, M.K., Centralized fog computing security platform for IoT and cloud in healthcare system, in: Fog Computing: Breakthroughs in Research and Practice, pp. 365–378, IGI global, 2018. 7. Sorel, R., Ahamed, S., II, Demartini, C., Conte, T.M., Liu, L., Claycomb, W.R., Takakura, H., 41st IEEE Annual Computer Software and Applications Conference. 2. Annual Computer Software and Applications Conference, IEEE Computer Society, 2017. 8. Chiang, M. and Zhang, T., Fog and IoT: An overview of research opportunities. IEEE Internet Things J., 3, 6, 854–864, 2016. Negash, B., Rahmani, A. M., Liljeberg, P., &Jantsch, A. (2018). Fog computing fundamentals in the internet-of-things. In Fog computing in the internet of things (pp. 3–13). Springer, Cham. 9. Atlam, H., Walters, R., Wills, G., Fog computing and the internet of things: a review, in: big data and cognitive computing, vol. 2, no. 2, pp. 10, 2018.

350  The Smart Cyber Ecosystem for Sustainable Development 10. Dastjerdi, A.V., Gupta, H., Calheiros, R.N., Ghosh, S.K., Buyya, R., Fog computing: Principles, architectures, and applications, in: Internet of Things, pp. 61–75, Morgan Kaufmann, 2016. 11. Firdhous, M., Ghazali, O., Hassan, S., Fog computing: Will it be the future of cloud computing? The Third International Conference on Informatics & Applications (ICIA2014), 2014. 12. Camhi, J., Former Cisco CEO John Chambers predicts 500 billion connected devices by 2025, in: Business Insider, 2015. 13. Pierson, R.M., How Does Fog Computing Differ from Edge Computing, 2017. Online: https:// readwrite. com/2016/08/05/fog-computing-different-edge- computing-pl1/. Accessed, 12. 14. Sorel, R., Ahamed, S., II, Demartini, C., Conte, T.M., Liu, L., Claycomb, W.R., Takakura, H., 41st IEEE Annual Computer Software and Applications Conference. 2. Annual Computer Software and Applications Conference, IEEE Computer Society, 2017. 15. Chiang, M. and Zhang, T., Fog and IoT: An overview of research opportunities. IEEE Internet Things J., 3, 6, 854–864, 2016. 16. Shropshire, J., Extending the cloud with fog: Security challenges & opportunities, 2014. 17. Bonomi, F., Milito, R., Zhu, J., Addepalli, S., Fog computing and its role in the internet of things. Proceedings of the first edition of the MCC workshop on Mobile cloud computing, ACM, pp. 13–16, 2012, August. 18. Stojmenovic, I. and Wen, S., The fog computing paradigm: Scenarios and security issues. 2014 Federated Conference on Computer Science and Information Systems, IEEE, pp. 1–8, 2014, September, Jaiswal, A. S., Thakare, V. M., &Sherekar, S. S. (2015). Performance based Analysis of Cloudlet Architectures in Mobile Cloud Computing. International Journal of Computer Applications, 975, 8887. 19. Pierson, R.M., How Does Fog Computing Differ from Edge Computing, 2017. Online: https:// readwrite. com/2016/08/05/fog-computing-different-edge- computing-pl1/. Accessed, 12. 20. Bahl, V., Emergence of micro datacenter (cloudlets/edges) for mobile computing. Microsoft Devices & Networking Summit 2015, 2015. 21. Sheth, V., Medium, 2019. https://medium.com/@virral/fog-edge-computing-and- cloudlets-c4db6cc9b15a. 22. Zhang, L., DZone, IoT Zone, 2018. https://dzone.com/articles/bringing-iot-to-the-cloud-fogcomputing-and-cloudl. 23. Naha, R.K., Garg, S., Georgakopoulos, D., Jayaraman, P.P., Gao, L., Xiang, Y., Ranjan, R., Fog Computing: Survey of trends, architectures, requirements, and research directions. IEEE access, 6, 47980–48009, 2018. 24. Solutions, C.F.C., Unleash the power of the Internet of Things, Cisco Systems Inc, 2015. 25. Dastjerdi, A.V., Gupta, H., Calheiros, R.N., Ghosh, S.K., Buyya, R., Fog computing: Principles, architectures, and applications, in: Internet of Things, pp. 61–75, 2016, Morgan Kaufmann. 26. Luan, T.H., Gao, L., Li, Z., Xiang, Y., Wei, G., Sun, L., Fog computing: Focusing on mobile users at the edge. arXiv preprint arXiv:1502.01815, 2015. 27. Shin, S.W. and Gu, G., Cloudwatcher: Network security monitoring using openflow in dynamic cloud networks, in: Network Protocols (ICNP) 2012, pp. 1–6, IEEE, 2012, October. 28. Kirat, D., Vigna, G., Kruegel, C., Barecloud: bare-metal analysis- based evasive malware detection, in: 23rd {USENIX} Security Symposium ({USENIX} Security 14), Pathan, A.S.K., Lee, H.W., Hong, C.S. (Eds.), pp. 287–301, 2014, 2006, February, Security in wireless sensor networks: issues and challenges. In 2006 8th International Conference Advanced Communication Technology (Vol. 2, pp. 6-pp). IEEE. 29. Kulkarni, S., Saha, S., Hockenbury, R., Preserving privacy in sensor-fog networks. The 9th International Conference for Internet Technology and Secured Transactions (ICITST-2014), IEEE, pp. 96–99, 2014, December.

Fog Computing Perspective  351 30. Stolfo, S.J., Salem, M.B., Keromytis, A.D., Fog computing: Mitigating insider data theft attacks in the cloud. 2012 IEEE symposium on security and privacy workshops, IEEE, pp. 125–128, 2012, May. 31. Dsouza, C., Ahn, G.J., Taguinod, M., Policy-driven security management for fog computing: Preliminary framework and a case study. Proceedings of the 2014 IEEE 15th International Conference on Information Reuse and Integration (IEEE IRI 2014), IEEE, pp. 16–23, 2014, August. 32. Lu, R., Liang, X., Li, X., Lin, X., Shen, X., Eppa: An efficient and privacy-preserving aggregation scheme for secure smart grid communications. IEEE Trans. Parallel Distrib. Syst., 23, 9, 1621–1631, 2012. 33. Zeng, L., Xu, S., Wang, Y., VMBackup: an efficient framework for online virtual machine image backup and recovery, in: Concurrency and Computation: Practice and Experience, vol. 28, no. 9, pp. 2630–2643, 2016. 34. Boumerdassi, S., Renault, É., Muhlethaler, P., A stateless time- based authenticated-message protocol for wireless sensor networks (stamp). 2016 IEEE Wireless Communications and Networking Conference, IEEE, pp. 1–6, 2016, April. 35. Lee, H., Lee, K., Shin, Y., Implementation and Performance Analysis of AES-128 CBC algorithm in WSNs. 2010 The 12th International Conference on Advanced Communication Technology (ICACT), vol. 1, IEEE, pp. 243–248, 2010, February. 36. Benzekki, K., El Fergougui, A., Elbelrhiti, E.A., A secure cloud computing architecture using homomorphic encryption. Int. J. Adv. Comput. Sci. Appl. (IJACSA), 7, 2, 293–298, 2016.

19 Cybersecurity and Privacy Fundamentals Ravi Verma

*

Department of CSE, Radharaman Institute of Technology & Science, Bhopal, India

Abstract

Cybersecurities consider different security aspects of the computing system involving communication with security and privacy mechanism. Cybersecurity is very wide to limited with few specific problems because it covers security and privacy of various applications, infrastructure, and system policies. Security is the main constraint for communication, where we need to protect our system from attackers which may bring many vulnerabilities for the communicating system; one needs to be ready with the kind of security mechanism to overcome this problems; cybersecurity is all about managing computing operation with reliability, security, and privacy; it makes this area more challenging and interesting for its researchers. As the advancement in computing technology and communication system is growing, it demands high level of communication security in the form of cybersecurity; the digital information is very difficult to manage from various cyber attackers and outside unauthorized users like eavesdroppers; cybersecurity considers all the security and privacy issues by adopting different infrastructure-based tools and techniques to manage all the security aspects of computing system; in addition, this technology introduces various new aspects of getting security at different services layer to fulfill different needs of the system. This chapter will present an impactful study about the fundamental of cybersecurity and privacy; in past, there are many research articles that have been published in the same area but we feel to design a new aspect for describing cybersecurity in more impactful manner so that it will be beneficial for many readers, researchers, academics, and students for learning about cybersecurity and privacy concerns. In this chapter, we investigate various types of attacks, malware, and viruses from their fundamental to find the weakness of system with existing tools and techniques; author will also like to concentrate on various privacy issues with its solutions along with various present and future challenges of cybersecurity operations. Keywords:  Cybersecurity, privacy, cryptography, cyber attacks, reliability, challenges of cybersecurity, malware, viruses

19.1 Introduction The aim of this chapter is to provide an impactful presentation and description to the area of cybersecurity which should be easier for the students to learn and those who study this Email: [email protected]; [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (353–378) © 2021 Scrivener Publishing LLC

353

354  The Smart Cyber Ecosystem for Sustainable Development area for the first time, and also helpful for the expert of the same area for further extension. This chapter will help students to learn the following topics. • • • • • • • • • • • •

Historical background and evolution of cyber crime Introduction to cybersecurity Classification of cyber crimes Reasons behind cyber crime Various types of cyber crime Various types of cyber attacks in information security Cybersecurity and privacy techniques Essential elements of cybersecurity Basic security concerns for cybersecurity Cybersecurity layered stack Basic security and privacy check list Future challenges of cybersecurity

We are not limited with these only, during explanation, if we felt the need to include more content which is the demand of this chapter, and definitely, we will try to cover all the other topics that are not mentioned in above list of objectives. Author also focused on kind of readers who read about this topic for the very first time; therefore, this chapter will be helpful for the fresh learners and readers, by introducing fundamental things of cybersecurity and privacy in day-to-day life of computing system.

19.2 Historical Background and Evolution of Cyber Crime The term cyber crime refers to all the illegal (criminal) activities, which is not according to cyber law and standards; such activities are using computer itself or other computing devices like phones, servers, and tablets, for targeting to perform criminal operations for personal benefits; it is done by all those people who want their individual financial benefits by hacking the financial data or due to some revenge. Internet technology comes into life around 1960s; that time, it is available for limited purpose, and only defense community, scientist, and researcher can have permission to access and work with Internet; that time, only physical damage is consider as a crime in computing technology, nothing logical things exist so one needs to think about the damage of infrastructure based only that can be easily handled; but in 1980s, as computing technology is getting popularity, damage is not limited to physical damage level but the logical damage has been introduced in the form of some computer malfunction and coding; such type of malicious functional was called virus; that was the first time when someone listen about the virus, such virus was less effective in Internet because Internet was only used by defense and research communities [1]. Around 1996, Internet was fully launched for public users; the magical Internet computing and operational services were quite impressive which make the Internet more and more popular in very less time; as a result, users rely on Internet services because it brings a computing environment which drastically changed the lives of people; now, Internet services are easily accessible via various graphical user interface (GUI) platforms which help its users in a friendly manner; now, many individuals

Cybersecurity and Privacy Fundamentals  355 are having knowledge of doing computing work through internet, transferring data from one end to other over Internet without worrying about the background complexity of doing so. As users increased, the challenges also got higher and higher; now, crime in computing technology was reflected by affecting millions individuals data and computers so that physical damage of computer as a crime was shifted to logical data manipulation for individual benefits. As defined in [2], at the same time, various types of attacks were found and rapidly increased as time passes; in a study, we found that, every second, there were approximately 25 to 30 computers that are affected by cyber attack. Until 2013, there were about 800 millions of internet users that were affected by the same; in similar investigation during 2011 to 2013, CERT India had released a report which mention approximately 308,350 different websites that have been hacked by someone else as an unauthorized accessing and controlling activities, due to such numerous attacks computing technology lost in millions per year.

19.3 Introduction to Cybersecurity Internet brings a magical computing platform for all its users and business applications; the 21st century will be the world of networking services which cover all most every area of business and life with full of automated, self-configured, self-healing, and self-computing technologies which provide a lot for people and business purpose with efficiency and consistency; but although we cannot ignore the fact that behind every successful technology, there is a big chain of problems; similarly, there are many vulnerability raised behind Internet services and communication technology due to cyber crime, to take care about such vulnerabilities Internet technology adapting the concept of cybersecurity which consider maximum vulnerabilities raised due to various types of security concerns as in [3]. Since networking technology allows many users and business processes to participate together in a process, to store, to access, and to transfer high volume of data which includes different category of data produced via sensitive and prioritize applications such data sets, which belong to commercial operational category and personal data category, all the commercial and personal data operations are going through a secure process to get objectives at the end of any transaction which brings trustworthy environment for business and users to rely on personalized data operations and, at the same time, security from cyber threats, attacks, viruses that are the big challenges in front of all the scientist and researchers of universe to protect our network and data from outside attackers and to immune the networking services from eavesdroppers. Now, as long as the internet services are getting expanded at the same time, one needs to expand the security measures for the well fair and future of Internet and digital computing world; the rapid growth of Internet services reflects the risk issues of digital communication as we knew; cybersecurity risk creates a big barriers for the economy of Internet; it requires a big investments in the direction of cybersecurity so that people can believe on digitalization and can smoothly participates in the commercial operations; it does high growth in Internet technology and addressing the solution for all the barriers of innovation, financial growth, and easy flow of information from anywhere and anytime.

356  The Smart Cyber Ecosystem for Sustainable Development Since Internet cover wide range of services via different elements of networking technologies and, at the same time, cybersecurity measures are also need to keep in parallel with this services architecture to get Internet services with security, reliability, and efficiency. At present, cybersecurity is one of the major areas of research because cyber crimes are increasing and rising challenges in front of many cybersecurity scientist and researchers; the advancement in cybersecurity technology will decide the future of Internet and its service platforms, and we can define cybersecurity like following manner. Definition: Cybersecurity is a part of Internet technology which works as a safe guard or protection layer to protect entire computers, devices, software, hardware, networks, and data as information to attempt unauthorized access or to attack from unauthorized users or eavesdroppers which aimed to misuse these data and network elements since cybersecurity is applicable for various things; therefore, it is also known as information security, application security, network security, and many others as defined in [8]. The following are the major areas need to be considered for the implementation of cybersecurity.

19.3.1 Application Security From the very beginning, when one decide to adopt some technique or methodology for the design and development of software system, one needs to consider software level security during the development phase which can protect application or software from threats which can affect the software during design, development, implementation, maintenance, and up-gradation phases according to [5]; basically, there are already few predefined procedure that have been deployed during application or software development which are as below. • • • •

Deployment of validation technique during the selection of input parameters. Deployment of authentication and authorization of different user’s module. Deployment of session and exception handling techniques. Deployment of audit trail and security standards.

19.3.2 Information Security Information security is very crucial task for Internet, to serve the data services with security and reliability, networking technology deploying various security checks to protect data from unauthorized access, and to avoid such activities during the transmission of data; the following are the major security mechanisms used by system as defined in [4]. • Deployment of encryption and decryption technique which aims to maintain security and privacy of data. • Deployment of user’s verification and validation techniques. • Deployment of firewall for the identification of users and network.

19.3.3 Recovery From Failure or Disaster Any system may crash any time due to human error or due to any natural disaster; therefore, data recovery from failure is very important consideration for cybersecurity because

Cybersecurity and Privacy Fundamentals  357 data may also get lost due to some attack; in this case, the key responsibility of cybersecurity expert is to resume the data operation activities as soon as possible. As defined in [18], to manage data recovery from disaster and failure system, the following majors should be adopted. • Deployment of data backup and recovery planning. • Deployment of risk assessment strategies from time to time. • Deployment of planning to handle disaster and recovery from it.

19.3.4 Network Security According to [6], network security covers various activities to protect our network because network can be easily hacked through some attacker or may get down due to some virus effects; therefore, cybersecurity mechanism includes mechanism to protect the reliability, security, privacy, integrity, and confidentiality of entire communication network; cybersecurity aims to avoid various threats and stopping them to enter and spread into the system; few of them are as follows. • • • •

Installation of anti-virus software. Installation of Firewall servers. Deployment of intrusion prevention and detection system. Deployment of virtual security services for managing security from remote location.

19.4 Classification of Cyber Crimes The cyber crime activities could be classified in internal cyber attacks or external cyber attacks based on the nature, and intention behind attack decides whether it is internal or external attack.

19.4.1 Internal Attacks Internal attackers in cyber crime can target the system of an authorized one in order to get access over the network for making crime. This is done due to any personal reason such as employee is not very much satisfied with the company or because of some revenge that may happen. As in [7], internal attacks could be performed by any employee of an organization because the attacker has accessing permissions so that he is very much aware about the policies, architecture, and security checks of organization; therefore, it is not very difficult task since it might be possible that any employee could be an attacker at any time in company so that every business has to take care about the confidentiality regarding process, policies, and security checks of organization to be kept in safe manner as much as possible, but although such vulnerability may raise any time or by anyone, therefore cybersecurity mechanism instructs to every organization to deploy internal intrusion detection system to get and identify the one who attacks on system.

358  The Smart Cyber Ecosystem for Sustainable Development

19.4.2 External Attacks When an organization, its computer, and network have been targeted through an external attacker which is not authorized by system but due to vulnerable design, development, and policies of system, an attacker gets confidential information through hacking strategies or steals the sensitive information through some insider to enter into the system as easy as possible. Sometimes, such attackers have been hired through someone, who wants to lose the reputation of the company in such an comparative environment of business world; due to external attack, many times, an organization may suffer from a big financially loss and likewise, at the same time, it may also lost client trust. As described in [9], external attacker keeps observing the logs generated by server; through this log detail, attacker can analyze the information to enter into the system and break system security parameters; due to poor design or development technique, many times, weaknesses of system have been identified by external attacker to target the system. Installation of firewall server at system boundaries may protect and help system to identify the attacker and block them before targeting the organization for unauthorized accessing.

19.4.3 Unstructured Attack Many times, attack is intentional but sometimes it may happen without any intention; when attack has been done by an immature, who just doing such things for a testing purpose and without aiming for any financial loss of organization, in some case, such immature one is doing this unknowingly just for fun or for the purpose of getting knowledge, they try to do this with a randomly chosen company; such types of cyber attacks comes into the category of unstructured attack as in [15].

19.4.4 Structured Attack This is the special category of attack which may be either internal or external attack based on the intention and planning of attacker. This category of attack is considering only highly skilled people having knowledge and experience of doing similar things; structured attack is a part of planning in which an experienced one will target an organization for their personal financial benefit; therefore, this is an intentional attack which already going on in mind of attacker. As in [14], once attackers decide their goal, they plan accordingly through some structure, since they have the capacity to access the entire network and have knowledge about the security tools and parameters, therefore they can bypass the internal intrusion detection system during attack and, at the end, they finished the crime successfully; such types of attackers are also called as professional cyber criminals; it is very dangerous for the entire communities, societies, and countries because structured attackers could be harmful for the entire nation as well.

19.5 Reasons Behind Cyber Crime Today’s business world competition among the participants and companies are on real trend, which encourage people for doing such criminal activities; some people believe on such crimes for making money and career; therefore, the following are the prominent reasons that are responsible for the growth of cyber crime.

Cybersecurity and Privacy Fundamentals  359

19.5.1 Making Money Many people believe that it is the easiest way to make money in very less time; so, expert attackers are hacking any commercial platform for making money; sometimes, they break security checks of users’ account, whereas, sometimes, they get ID (identification number) and PWD (password) of users via hacking skills to get money.

19.5.2 Gaining Financial Growth and Reputation Many business organizations are hiring skilled attackers, aiming to get down other business financial growth and reputation in market because they feel this is the more prominent way to reduce number of clients and customer of any organization; by doing this practice, they feel that they will connect more clients in business and get more popularity in business with financial growth as well.

19.5.3 Revenge Few people are committing cyber crime, aiming to take revenge with other employee of company or business; therefore, they are planning to attempt this for defaming organization reputation and financial loss.

19.5.4 For Making Fun Many people are committing this crime for just making fun with other; sometimes, they just want to test and experiment their code or tool without any criminal intention, and due to immaturity, they do not realize the after effect of this practice.

19.5.5 To Recognize Few people are claiming that hacking the most secured network is the easiest and most challenging way to get famous and recognized in very less time; therefore, they try and think about to crack and hack the highly secured computer, servers, or networks like very confidential and highly secured military network or defense site.

19.5.6 Business Analysis and Decision Making Cyber crime is sometimes used as a tool of boniness analysis, planning, and decisionmaking; by attempting internal attack, any organization can know the policies of other organization which helps one to know in advance about planning of production and future decision of business for maximum growth and higher financial benefits as in [17].

19.6 Various Types of Cyber Crime There are a large number of cyber crimes that have been identified in network; few of them are as follows.

360  The Smart Cyber Ecosystem for Sustainable Development

19.6.1 Cyber Stalking This is the medium of acting criminal activities such as stalking or blackmailing people through computer or other communication devices; the aim behind such crime is to defame someone via internet services like using email, messenger, and other web-based activities to expose one sexually or personally. Cyber stalking is also considered as a crime where attackers steal the personal information like images, videos, and other confidential data of someone, aiming to misuse it or selling it to make money.

19.6.2 Sexual Harassment or Child Pornography Today’s internet users are having heavy engagement with internet through social media and communication apps; now, people are keeping their personal data over cloud environment; in this case, such openly available data can be hacked by any attacker, aiming to sexually harass one or may use it for pornography.

19.6.3 Forgery Nowadays, skilled experts are involved in forgery activities; to do this, they are preparing fake documents of property using digital technology which looks like the original one so that no one can identify it as a fake document; such document has been used for selling property to someone without presence of property’s original owner; criminals are very active to do such forgery using digital technology as discussed in [10].

19.6.4 Crime Related to Privacy of Software and Network Resources Illegal uses of network resources and freely available software download are consider as a source of crime because there is a license policy launched with software; if any business or person like to use that software or network resources, they have to get license from authorized dealer to use software. The licensed software is the reliable one for installation in computer without losing security concerns; therefore, if one buys it in a reliable way, they got software with best security features and services which protect it from virus and attack as in [13].

19.6.5 Cyber Terrorism Expert attacker could find the way to break highly secured commercial or military sever for fetching confidential information for committing cyber terrorism attack, aiming for financial benefits or harming people of our national by helping terrorist to plan attack or to exchange confidential information with terrorist.

19.6.6 Phishing, Vishing, and Smishing As the technology is growing, types of committing crimes are also growing with more prominent way like phishing, where an attacker send a mail to targeted user when users open such mail or clicking on any highlighted link, attack becomes active to steal money

Cybersecurity and Privacy Fundamentals  361 from users account instantly; similar types of crime when done through voice messages on mobile or by sending some text message services will be consider as Vishing and Smishing; by doing this practice, one can easily got user ID, Password, or Credit/Debit Card details in easiest way to control over the user commercial or financial account as discussed in [11].

19.6.7 Malfunction It is the technique of deploying malicious code, aiming to destroy any physical computing resources such as storage devices, communication device, and computer. Malfunctions consider as virus which can destroy the entire computing resources in very less time as in [8].

19.6.8 Server Hacking Hacking is a technical approach to make control over the computing resources like server and other computing resources, aiming to misuse it for personal purpose or to modify and destroy information at server site. Nowadays, industries are hiring such experts who are having specialization in hacking the computer, web site, or server. As in [19], hacker is a highly skilled person having enough technical knowledge; sometimes, hacker protects the organization server by fixing security checks and server vulnerabilities.

19.6.9 Spreading Virus Virus is a kind of malicious code which can enter into the computer system via Internet and other communication devices; virus can crash the information at server and has capacity to damage computer devices, aiming for financial loss of an organization. Virus is also created by a technically sound experts; through virus, one can get control over other system; virus can be spread into millions of computers of any network in few seconds.

19.6.10 Spamming, Cross Site Scripting, and Web Jacking Sometimes, Internet users also face various types of attacks in the form of spamming mailing activities, where an unknown sender will mail a unsolicited mail to someone so that when user open such mail, the embedded code will spread into the system as a virus for crashing users’ data for fetching some confidential information; such all mails will be considered as spam mail messages. Similarly, using cross site scripting techniques, an attacker can inject a malicious code into the targeted computer; it may be a server to get and send sensitive confidential information to remote system for making some financial or other crime; with the help of web jacking technique, an attacker can get control over the web site of targeted organization as defined in [12]; in this case, hijacker may modify the contents of the web site for solving their purpose.

19.7 Various Types of Cyber Attacks in Information Security According to [20], the digital technology is growing its lead to increase cyber crime in large scale; therefore, we can see today’s network are facing many problems due to

362  The Smart Cyber Ecosystem for Sustainable Development such attacks and viruses because many unethical and illegal practices are going on in information security which we called as network attacks, viruses, and worms. As in [24, 27], attacks and viruses target to exploit the information and to modify and steal sensitive data for the financial benefit of an individual person; actually, these attacks are spreading into the computer or network in the form of malicious code which execute in system automatically without permission; it is a very serious challenge to protect network and computer from such attacks because today most of the people like to work in digital platform, and due to this dependency, security challenges are growing higher and higher; in order to understand the concept of entrance, working, and after side effects of various types of information security attacks, we can classify this into following major categories.

19.7.1 Web-Based Attacks in Information Security According to [31], this category of attacks considers all the different types of attacks which occurred on a web platform like any web site or web-based applications. Figure 19.1 describes some basic web-based attacks. i.

Injecting Attacks: An injecting attack has been used to fetch or alter sensitive information from some targeted web sites or applications; to do this, malicious data has been injected to the web server or application which spread automatically to alter data values and steal required information; examples for injecting attack are SQL database injection, XML server injection, log trail injection, etc. ii. Spoofing: It is a type of hacking, through which an attacker can target to any Domain Name Server (DNS) in order to identify any Internet address or to divert Internet addressing traffic in a direction to other unauthorized server, computer, or any other intruder to capture different confidential IP address for the purpose of criminal activities or other misuse. This category of attack may lead for very serious problem in digital communication system; such attack is not easy to detect through other attack detection networking techniques as in [22]. iii. Hijacking a Session Object: In a protected network, two or more users have been communicating to each other via a secured connection protocol; that time, an attacker can hijack a user session so that any outside unauthorized user can access all the data which has been communicated through this session and this data can be misuse by an attacker in near future. According

Web Based Attacks

Injectings

Spoofing

Hijacking

Phishing

Figure 19.1  Various types of web based attacks.

Brute force

DoS

Universal Resource Encoding

File Inclusion

Man in the Middle

Cybersecurity and Privacy Fundamentals  363 to [26], this attack captures and steals the cookies saved by web application for efficient web operation, and through, these cookies any attacker can get hijack a session over protected network. iv. Phishing: Through this attack, an attacker may steal sensitive information of any user like credit card details and commercial banking login detail to use this information for personal financial benefits. To adopt this, an attacker just tries to masquerading himself as an original trustworthy object of communication world as in [23, 28]. v. Brute Force Attack: This particular category of attack is putting its full concentration on hit and trial method for fetching sensitive information like ID and Password, an attacker guessing the different ID and Password to crack the security checks; this attack especially used for cracking encrypted data security in order to identify encryption pattern to decrypt information in actual form. vi. DoS (Denial-of-Service) Attack: This is a special category of attack through which an attacker could be able to control over the targeted server, and once an intruder does this successfully, this server is not available anymore for the actual authorized users; similarly, same practice can also be possible for any computing resources as in [18, 21]. For doing this thing, attacker uses the concept of flooding with a system and single internet connection to targeting any server. This attack can target any server with the following purposes. vii. Dos for Bandwidth Control and Observation: Aim behind such attack is to control over the bandwidth of any targeted server or it may be an organizational site to observe the operational bandwidth in bits per second. viii. DoS Protocol Attack: Aim behind this attack is to consume all the available computing resources; it may be any peripherals of network, a computer, or a server. ix. DoS Attack on Application layer: Through this attack, an attacker can crash any web server or any web-based application layer applications or services. x. Universal Resource encoding: This attack targeting an URL (Universal Resource Locator) or any part of it, aiming to copy it and making our own server for designing and delivering web contents or pages like official web site for making any financial benefits as in [21]. xi. File Inclusion Attacks: This attack is working as a virus and has spreading nature, so that once it enters into the file system for unauthorized access, it will spread over all the file of storage to get it infected via malicious file codes. xii. Man-in-the-Middle Attack: As names define that this attack is acting as a middle man for unauthorized accessing among two authorized objects of network. It is a type of attacks which works in invisible way to enter and intercept the connectivity between client and server by creating a virtual bridge for unauthorized accessing of data for making insertion, deletion, and alteration operation over data values through connection as in [29].

364  The Smart Cyber Ecosystem for Sustainable Development

19.7.2 System-Based Attacks in Information Security This category of attack is considering all the things which affect individual system like computer or a dedicated network. The goals behind this attack are to crashing a file system of any computer or server directory structure for compromising computer storage security. There are few common categories of such attacks, which are described in Figure 19.2; all such attacks are very harmful for any system. i.

ii.

iii.

iv.

v.

Virus: It is a category of malicious programs sometimes available in freeware on Internet as freeware software; this is a very serious type of attacks which can spread into numerous computer through Internet connection to infect all; once a virus when entered into files system, it can have capacity for self-replication; without knowledge of actual users, such replication not only engages storage space but also damages the storage capacity of computer which we called bad sector portion of storage device; the execution of such virus can also harms sensitive information file by crashing it forever as in [25]. Worms: It is consider as a malware attack; this works similar like virus like it is also self-replicated computer programs which can spread into many computers for unauthorized accessing and crashing files; the source of this attack is email attachments or via some Internet advertisement hyperlinks; when any user receive and download such attachments or click on hyperlinks, worms will enter into the system and then automatically execute and spread into the computers as according to [30]. Trojan Horse: The invisible malicious code which intent to change computer setting and sometimes perform unusual activities, such attack occurred unexpectedly and always works as background application; the system user cannot understand the intents of this attack because it will mislead to user about its intent; it is working blindly for harming computer system. Backdoors: As name suggests that a type of attacks which can bypass the authentication process to enter into the system, therefore it is called as backdoor entry attack; this attack developed through highly skilled computer experts for resolving many computing purpose or performing troubleshooting operation when required. Booting Virus: Booting is a terminology used computer science which is related with the starting of computer machine; therefore, this category of virus is working on the concept of computer system booting process; when system gets on this, virus will be active for infecting data which is stored

System Based Attacks

Virus

Worms

Trojan Horse

Back doors

Booting Virus

Figure 19.2  Various types of system based attacks.

Polymorp hic Virus

Macros

Spyware

Stealth Virus

Logic Bomb

Cybersecurity and Privacy Fundamentals  365 inside hard disk drive; similarly, this is also harmful for different parts of storage devices like master boot record. For example, crazy boot virus is designed for infecting Microsoft Windows Operating System as in [22]. vi. Polymorphic Virus: A category of virus which has self-replicating capability with specialty of making fool to virus detection software like antivirus by changing virus signatures every time whenever it gets replicated in new form, variation in virus signature through some encryption technology brings difficulties for virus detection software. vii. Macros: This virus is especially designed for infecting the files having extension like .doc, .docx, and .xlsx; when virus infects these files, it may be completely lost; therefore, it is very harmful for confidential files. Through Internet, it can infect one file to other file of different computer; Melissa and Bloodhound are some examples. viii. Spyware: This is a personal information hacker which can save your sensitive information and send it back to some attacker’s website; spyware is installed automatically without knowledge of user when user try to install any free software from the Internet as in [18]. ix. Stealth Virus: This virus is similar like Polymorphic Virus; it follows hidden strategy that means it changes its appearance every time so that it cannot be identified as a virus through any antivirus and virus detection technique; for example, Frodo is a stealth virus. x. Logic Bomb: This is a time base attack which can remain in computer for long period of time and execute on any predesigned event for a particular date or time; such an event based attacking is called logic bomb as in [32].

19.8 Cybersecurity and Privacy Techniques As the Internet technology is growing with higher number of Internet users at the same time attackers, hijackers, and viruses also gets higher in large number, now, all the services of Internet are fully captured through all around availability of different types of attackers; to maintain the trust of users, it is the key responsibility of cyber experts to deal with more advance and secured technology which motivates Internet users to use Internet services platform without worry; cybersecurity is the challenging area in front of all the researcher scientist and experts; therefore, today, digital platform is bringing more secure services to its user with time-to-time availability of more advance and improved version of services as in [33]. The following are the major techniques used to counter the attacks.

19.8.1 Authentication and Authorization It is the technique which perform verification and validation of an individual one and ensure that the identified person is valid or not to perform some task; usually, every application, web site, and other digital platform always go through a process of authentication and authorization via allowing users to enter user ID and Password; when someone successfully entered into the system using the process of authentication, it defines that the specific users is authorized by system to do further operations, but in large system, there are many types

366  The Smart Cyber Ecosystem for Sustainable Development users that have been created based on their accessing permission and role, so that after the process of authentication, authorization has been taken place, which defines that the authenticated user is only authorized for their specific accessing permission. According to [34], nowadays, cyber crimes are increasing in day-to-day business services; therefore, an additional feature of two-step authentication has been added, called One-Time Password (OTP); according to this, it is generated for single purpose only and its having limited life time to use in process, which means every time an OTP has been generated whenever system call for authentication. In this process, an OTP has been sent to user’s registered mobile number or mail which was specified during the process of registration; such process brings more secure way of authentication because now user needs to have three things like ID, Password, and OTP to get verified and entered into the system. As the advancement is growing, artificial intelligence and machine learning brings a more advance process of two-step authentication through bio-metric verification like finger prints matching and authentication through eye retina image or via other parts of human body.

19.8.2 Cryptography It is the way of secret writing which has been proposed by aiming to maintain data integrity, confidentiality, and privacy of data from unauthorized attackers; since data is openly transmitted from one to other end over different transmission media, during this, anyone can hack such confidential information for making money. Cryptography is a technique to transmit such confidential information more securely using some data encoding and decoding techniques so that original data to be converted in some unreadable format to hacker which cannot be useful for anyone or any purpose. Internet techniques do this practice before transmitting the original data to other sources; the actual receiver has a key to convert into original form at sources, but in between this, Internet encryption mechanism ensures the integrity, consistency, reliability, confidentiality, and authenticity of data transmission that form actual sender to actual receiver as in [29]. An encryption technique uses many types of symmetric and asymmetric algorithm to lock such confidential information such as symmetric and asymmetric encryption algorithm, every algorithm having their own mathematical construct to convert the data into secret code. Figure 19.3 describes the actual procedure and sequence of pattern to perform encryption techniques among sender and receiver system.

SENDER

Plain Text

Encryprtion of Plain Text with Key

RECEIVER

Transmission Media

Cipher Text

Figure 19.3  Process of cryptography between sender and receiver.

Plain Text

Decryption of Cipher Text with Key

Cybersecurity and Privacy Fundamentals  367

19.8.2.1 Symmetric Key Encryption In this approach, data is encoded and decoded using same key at both sender and receiver end; this key has been shared between sender and receiver for the purpose of locking and unlocking the data before sending and after receiving data. In symmetric key cryptography, a signal key sharing is the complex task which a sender needs to share with receiver via telephone calling or some other safest way that needs to be use because, if this key will be hacked, then security cannot be maintain; therefore to eliminate weakness of symmetric cryptography, one can go with asymmetric cryptography techniques.

19.8.2.2 Asymmetric Key Encryption It is more suitable approach for encryption where the concept of two keys has been introduced; one is for sender for encryption of data, whereas other is for receiver for decrypt the data. These two keys are also called as public and private key as the name suggest that public key is the key which well-known key for all, whereas the private key is only the key used for some particular individual.

19.8.3 Installation of Antivirus Internet technology is now working as a most important part of business and people life; as the users are increasing, there are various types of malicious codes like virus, attacks, worms, and malicious function like computer programs that are also targeting the Internet resources. According to its design specifications, such malicious code can spread over the Internet in very less time to many users for breaking security parameters by destroying the data which is store inside server or to make any financial benefits by hacking sensitive information like ID and password. In order to protect Internet and its users from such malicious codes to enter into the system, computer software developers and designer have recommended installing a programs called antivirus, which can protect our system from such malicious code as in [34]. Antivirus continuously observes and fights against virus and performs various operations like to identify the virus and clean it from the system; it means an antivirus is protecting our system from entering new virus and cleaning the virus which is already in system. In internet, there are lots of virus that have been introduced everyday; therefore, one needs to update antivirus database from time to time to protect our system and to fight against these worms, virus, and attacks; today, various types of antivirus are available in market like Net protector, Quick Heal, and many others.

19.8.4 Digital Signature Digital signature is a technique to manage security of data over Internet by adopting various validation and verification techniques in order to validate the contents of any document, which have been sent by someone and also used for authenticating to associate users. Digital signature is validating the data by adopting the concept of data integrity; to perform verification, the digital signature has been made by encrypting the data at

368  The Smart Cyber Ecosystem for Sustainable Development sender’s end using private key, such encrypted data has been encapsulated with original data which has been transmitted over Internet to its receiver; now, at the receiver’s end, receiver decrypts it by pubic key of sender to find original data; at the end, the received data will be compared with original data; if both are the same, then it means that data has safely reached to its desired receiver without any tampering in between transmission; at the same time, this process also authenticates the sender because data is decrypting with the public key of sender and data is matching, which means that sender has verified and data integrity has been achieved as in [36]. Nowadays, as Internet technology is growing, therefore more and more data as file and other means have been transmitted over Internet with high volume; in that case, to manage data integrity with security, digital signature is the essential technique and primary need of today’s financial transitions. Digital signature is not only important for doing validation and verification of content of document and authentication of users but also useful for avoiding many situations where a fraud can be claimed through digital transition. Figure 19.4 describes the process of digital signature in cybersecurity.

Hash Code

Original Data 10101010101010101 01010101010100101 01010101011111111 11111111111111111 11111100000000000 00010101010101100 10101010101010101

111111111010101010101 010101010111111111100 100000000001010101010 010101001010101010100 000000000000000001111 111111111111100010110 111000111000001111111 Implementing Hash Algorithm Encryted with Hash using Private Key

Certificate Attachement og Certificate with Digital Signature

10101010101010101 01010101010100101 01010101011111111 11111111111111111 11111100000000000 Digital Signature

11101010110101010101 01010101010101010110 11010101010101010101 01010101010010101010 10100101010101001010 Digitally Signed Data

Figure 19.4  Process of digital signature in cybersecurity.

Cybersecurity and Privacy Fundamentals  369

internet Internet/Wide Area Network

Firewall

Local Network/Tacranet

Figure 19.5  Installation of firewall between internet and intranet network.

19.8.5 Firewall Firewall is a special type of security device especially designed to protect an individual organizational network from different threats, virus, and attacks. It is a combination of hardware and software which act as a safe guard or protection layer between an individual network and Internet. Firewall provides an efficient way to filter all the incoming and outgoing access of an individual network through firewall; one can set limit of users who can send information to outside world and can send information to own network. Today, many organizations demands high security and safely while using Internet services, firewall especially designed for such organization; sometimes, organization needs to protect Internet protocol addresses along with port address for different application services which can be implemented via firewall with the concept of traffic management with security in which it provides a way to configure and monitor different traffic of port as organizations deal with inbound and outbound traffic system; therefore, if firewall correctly configured from time to time, then it ensures that only authorized nodes can use the ports for accessing and remaining all can be blacklisted to avoid all unknown sources to interact with system, as defined in Figure 19.5, which describes how firewall protects our system from outside unauthorized users like hackers and eavesdroppers. Nowadays, in many operating system like windows 2003, 2008, ME, NT, and professionals having their own pre-embedded firewall, user needs to manually configure it as their need, based on the manual configuration firewall designed rules and regulations to protect individual system or network from outside attackers as in [19]. In conclusion, we can say that firewall is essential for every organization and every internet user because the present of firewall is not only protecting the system or server from virus and worms but also it avoids an attacker to attack on network infrastructure for launching an attack like DoS attack.

19.8.6 Steganography It is another important technique to get cybersecurity while using Internet services; using steganography, one can hide secret information within a document file or any other means like image file and program; in this technique, secret communication can be achieved because only involved users can have the information about embedded message and such

370  The Smart Cyber Ecosystem for Sustainable Development message is not visible to others because it is mixed with some other image or document file and can find using a special software only. The big advantage of this technique is an attacker cannot easily identify the file having embedded secret message, only the authorized sender and receiver having permission to use special software to transmit and find original message. To know how secret information has been mixed with an image, we use an example like an image file having many pixels and every pixel is represented by 24 bits, if the last 3 bits of pixel has been altered for hiding secret information, in this case, the resultant image will have an unnoticeable changes compared to the previous quality of that image such changes can be detect by only an experience one; in this way, we can transmit a secret information by some invisible mode of communication same operation of hiding secret information can also be possible with audio or video files, to use such services various types of software are available in market like OpenStego and Xiao.

19.9 Essential Elements of Cybersecurity Before we discuss about cyber crime, we need to go through with elements of security which bring security consideration and provide reliable trust worthy digital environment to its client, whenever any transition will be initiated and completed while keeping all the following elements as defined in Figure 19.6, always consider as secured, reliable, and transparent category of transition. i. Availability Availability ensures that weather any information or computing resources is available whenever it is requested by some user. Availability also ensures the accountability factors because a secure network always demand to account the security measures continuously throughout the operation going on; therefore, tracking of hacker location and tracing of unauthorized users are crucial task for any expert that can be only possible when such accountability measure keep available continuously as in [27]. ii Integrity Integrity ensures that all the data and resources accessed is legitimate, accurate, and protected from outside eavesdropper or we can, say, hackers, so that computing operations are going to be processed on actual data values, to bring reliability and security in digital communication platform data integrity plays an crucial role to its users so that users can believe on real-time computing

Confidentility

Integrity

Availability

Authenticity

Figure 19.6  Essential elements of cybersecurity.

NonRepudiation

Cybersecurity and Privacy Fundamentals  371 operation and can participate smoothly for commercial operation anytime and from anywhere. To ensure integrity of data network that uses hamming code, check sum computation and cyclic code redundancy control techniques. iii. Authenticity Authenticity is also an essential element of cybersecurity; it is a process of confirming and ensuring that a user is identified as an authorized one and he/ she is genuine to access computing resources according to permission privileges; in other way, we can also say that authenticity is a process of verification of users; normally, many organization will adopt such process by providing login or via some bio metric module to access data and resources so that whenever user try to access anything they have to finish login or bio-metric process successfully as in [17]. iv. Confidentiality Confidentiality ensures that only the authorized users can have permission to access all the sensitive information. It takes care about the uses of confidential information as well as other computing resources that should be only accessible via authorized one. A confidential communication brings reliability among the users of any commercial or non-commercial organizations; confidentiality can be achieved by adopting various permission privileges to different users for accessing information and other resources. v. Non-Repudiation This technique ensures about data integrity issues that means data should reach to its authorized receiver as it is sanded through sender that the communication has been taken place between two or more authorized users via digital signature and encryption-based algorithm techniques to secure sensitive information from attacker or hackers. This feature of security provides a way for secret writing so that, apart from authorized user, no one can be able to read or understand the things; whenever data has been transmitted via digital signature or encryption based technique, it brings transparency, security, consistency, and authenticity to its users.

19.10 Basic Security Concerns for Cybersecurity Nowadays, digital communication system cover a lots of basic needs of human beings; there are various application service platforms gathering variety of Internet services for its user in efficient and easy manner; people like to go with digital commercial transition because it saves time and on demand efficient services and provides fast solution for everything on finger tips; as a result, billions of people are using Internet applications and services 24*7; this brings big challenges in front of all the security manager and inventors to manage cybersecurity at every part of network; in near future, more and more business and people will depend on such digital platform; that time, it is not easy to manage, control, and protect our network from attackers, viruses, and worms; therefore, one needs to design and develop more advance standards, protocols, and guideline for managing safety and privacy of sensitive things from outside unauthorized users, but apart from these, general awareness among the Internet and application users also plays an important role as a best solution

372  The Smart Cyber Ecosystem for Sustainable Development

Precaution

Cybersecurity & Privacy Consideration Reaction

Maintenance

Figure 19.7  Cybersecurity concern.

strategy for making personal information and computing resources immune from viruses and attacks as discussed in [30]. Most of the virus, attacks, and worms need some human error or mistake to extend its effect on network; therefore, Figure 19.7 describes the basic things to be considered to account cybersecurity and privacy issues.

19.10.1 Precaution It considers all the preventive measures at application, software, and hardware level before participating in a process of digital communication, which means in advance stage, users have to take measures of security in the form of installing antivirus application software, configuring firewall services, and keeping security management guideline awareness during the use of Internet services like mailing services, commercial baking, transition, and payment services. According to a latest report related on cyber crime, there are 90% of crimes that are done because of the lack of awareness of people which give a chance to an attackers or hacker to attempt an attack for hacking sensitive information for making money for personal benefits; similarly, variety of crime could be made possible. To avoid such security concern one needs to be aware enough for the best uses of Internet services and computing resources as per the guideline and standards define by respective authority of cyber crime control organizations as in [24,25].

19.10.2 Maintenance Maintenance is one of the most important things, which every application or Internet user needs to keep in mind because timely maintenance is a lifeline of security measures; in digital environment, everything goes on in real-time environment which means attacker and hackers always try to break system and application security in active mode; therefore, to face such a crucial attack in the form of virus, worms, and eavesdroppers, user’s system software also needs to be fully maintain and up to date to handle, protect, and avoid from malicious code. User needs to follow few basic practices as a part of digital life like antivirus

Cybersecurity and Privacy Fundamentals  373 needs to be update on time, only buy legal software which needs to be update from time to time with new version and security features, and do not use openly available free wireless communication channel for accessing services. In conclusion, we can say a secured networking service always comes in parallel with regular maintenance activities.

19.10.3 Reactions Reactions consider all the process and activities which need to be taken after detecting any criminal incidents; this is always challenging in digital communication environment to take security measures that how to deal with criminals, attackers, and viruses on real-time business process.

19.11 Cybersecurity Layered Stack According to [29, 31], cybersecurity needs to cover all the area of digital communication; there are so many things which need to be protect to avoid attack and virus, since skilled attacker always concentrate on a single weakness of Internet services, therefore we need to follow different types of security concerns so that we can avoid the chances of unauthorized accessing of information and computing resources. Internet involves various inter-process communications from the beginning of initiating a process to requested intermediate and next to targeted server throughout the completion of process with some output; during this, various layers have been involved that have been described in Table 19.1, which represents a layer-wise functional structure to manage cybersecurity in network. Table 19.1  Layered structure of cybersecurity. Layers

Description

Security measures

05

User’s Security Layer

This layer considers all the security measure to verify all the users through authentication like login or bio metric verification module.

04

Application Security Layer

This layer covers all the security standards used for managing security at application level which considers only legal software application with up to date maintenance of software from time to time.

03

System Security Layer

This layer considers system security guideline which cares about the essential need of system security like always work with antivirus application, firewall should be on, and always follow authorization to avoid unauthorized access.

02

Network Security Layer

This layer covers the security standards to protect network and its resources from unauthorized accessing, attacks, and threats.

01

Physical Security layer

This considers safeguards for all the physical security like protecting ID, password, etc.

374  The Smart Cyber Ecosystem for Sustainable Development

19.12 Basic Security and Privacy Check List Technological advancement and invention is the responsibility of scientist and computer developers, but apart from these invention and advancement, every user of computer and Internet have to follow some guideline to use computer and Internet services; a basic awareness regarding to use computing services can protect one from big loss; the proper use of technology is always fruitful to its users; if user follows the below check list, then, definitely, cyber crime will be reduced in high numbers as defined in [32]. • Use strong password and other credential like credit card pin and related details. • Always use a system which is fully protected with licensed anti-virus software. • Always keep operating system and applications updated and up to date with latest version. • Keep your confidential and sensitive data in separate backup regularly. • During communication always use strong crypto-graphical technique. • Always protect confidential documents through digital signature. • Make your system firewall on every time. • Always follow the protocols and guideline of Internet and application uses. • Do not use freely available Internet connection like open free Wi-Fi for browsing or doing transaction activities. • Do not click on unknown hyperlinks, advertisements, and keep your browser ad blocker on every time. • Always download any application or document through well-known and reliable sources.

19.13 Future Challenges of Cybersecurity Cybersecurity is one of the most crucial technologies and requirements for future business for overall nations and countries, since reliable and secured service architecture is only the backbone of Internet and digital communication system, as we knew that already a big number of business and people have been dependent on communication digital technology and the result is awesome; now, integration of most of the digital platform with IoT technology brings a easiest way of communication; now, people like to be a part of such a magical digital communication; this victim of things that is now digital platform is growingly adopting the trust of business and people, to achieve this result of trustworthiness; there is a big layer of security that has been implemented which covers many vulnerabilities of present and future business and challenges as discussed in [12, 18]. Behind every successful technology, there is a big chain of challenge that is always there which creates many barriers for breaking belief and trust of people by attacking the system security at application, infrastructure, software, hardware, and any other parts of communication and network; cyber crimes are growing every day; now, young generation takes this area as a challenge for making their benefits to break system security through hacking, hijacking, and spreading malicious codes to gain financial benefits. Now, people believe to participate in criminal activities related to cyber system because they want big profit in less time and effort,

Cybersecurity and Privacy Fundamentals  375 and such things encourage many computer virus developers and hackers to break security system of confidential servers or to create new viruses for application for making money. According to the present scenario of COVID-19 infection, we look forward and imagine the future possibilities of many business applications that many organizations are facing financial loss during financial year 2020–2021 because they are not on digital platform; according to present lockdown situation, the production and delivery of many business have been affected due to COVID-19 effect; therefore, many people learn so many things and planning to structure their business on digital platform, so that in future, they will not get affected like present. This report reflects the image of future growth of Internet services and demands more secure and reliable era of communication, and looking forward to this, the following are the major challenging areas for cybersecurity and privacy technology. • Designing and developing advanced secured data storage platform so that data can be kept securely during, before, and after communication. We need a more secure technology which can solve the problems occurred due to “Ransomware Attack”. It is a big challenge for IT professionals and cyber­ security experts to provide safe guard from this attack in future. • Block-chain technology is the most important technologies for future Internet services; this technology brings a smooth way to exchange different currency and deal internationally according to business need; this area will be very challenging for future of cybersecurity because crime related to currency like Bitcoin has growing day to day; therefore, cybersecurity experts need to focus on this area of challenge. • As “Internet of Things” technology is emerging with various Internet and digital services platforms for making people and business practices as easy as possible, according to the concept of IoT, various living and nonliving objects of day-to-day life are getting connected to accomplish any task without human to human interaction; this may lead the challenges like any attack can hack such devices to control over various IoT operations and fetch of sensitive confidential information. • Expansion of Artificial Intelligence technology will help to business and network to identify and detect attacks on time so that timely decision has been taken place, and thus, we can protect our business and network from possible losses; therefore, machine learning tools and integration of big data technology through Artificial Intelligence technology (AI) expansion will make cybersecurity system stronger and will be a big protector for future business planning, strategies, and policies. • The cloud computing is the greatest inventions in computing era; therefore, attackers are targeting to spread any virus or to break security system so that all the associated user’s personal data like sensitive credentials will be accessed unauthorized and one can easily use this information for any kind of loss. Since cloud is an open system architecture where data object are openly available for its users, this is the key responsibility of cybersecurity experts that is to maintain high secured system which is immune enough to protect and manage the complete elements related to data and computing resources like server.

376  The Smart Cyber Ecosystem for Sustainable Development

References 1. Anderson, J.P., Computer security threat monitoring and surveillance, p. 79F296400, James P. Anderson Company, Philadelphia, PA, USA, Tech. Rep, Apr. 1980. 2. Comer, D.E. and Droms, R.E., Computer Networks and Internets, Prentice Hall, 2003. 3. Cárdenas, A. et al., Research challenges for the security of control systems, HotSec., Association for Computing Machinery (ACM), Berkeley, CA, USA, pp. 1–6, 2008. 4. Lee, A., Cyber physical systems: Design challenges. Proc. 11th IEEE Int. Symp. Object Compon.Oriented Real-Time Distrib. Comput. (ISORC), Orlando, FL, USA, p. 363_369, May 2008. 5. Shi, J. et al., A survey of cyber-physical systems. Proc. Int. Conf. Wireless Commun. Signal Process. (WCSP), p. 1_6, 2011. 6. Lee, M.G., Securing the human to protect the system: Human factors in cybersecurity. Proc. 7th IET Int. Conf. Syst. Saf., Cyber Secur. Conf., p. 41, 2012. 7. Zou, C., Zheng, Z., Liu, Z. et al., Application of cybersecurity in industrial control systems based on security protection technology for electrical secondary system. Power Syst. Technol., 37, 11, 3227–3232, Nov. 2013. 8. Srikantha, P. and Kundur, D., Denial of service attacks and mitigation for stability in cyber-enabled power grid. Proc. Innovative Smart Grid Technologies Conf. (ISGT), 2015 IEEE Power Energy Society, pp. 1–5, 2015. 9. Cuadra, L., Salcedo-Sanz, S., Del Ser, J. et al., A critical review of robustness in power grids using complex networks concepts. Energies, 8, 9, 9211–9265, 2015. 10. zhun, A. et al., The Industrial Control Systems Cyber Emergency Response Team (ICS-CERT), in: Cyber-attack against Ukrainian critical infrastructure, Alert (IR-ALERTH- 16-056-01), 2016. 11. Tang, Y., Chen, Q., Li, M. et al., Overview on cyber-attacks against cyber physical power system. Autom. Electr. Power Syst., 40, 17, 59–69, Sept. 2016. 12. Moreira, N., Molina, E., Lázaro, J. et al., Cybersecurity in substation automation systems. Renew. Sustain. Energy Rev., 54, 1552–1562, 2016. 13. Evancich, N. and Li, J., Attacks on industrial control systems, in: Cybersecurity of SCADA and Other Industrial Control Systems, vol. 6, p. 95_110, Springer International Publishing Switzerland, 2016. 14. Yan, J., He, H., Zhong, X. et al., Q-learning based vulnerability analysis of smart grid against sequential topology attacks. IEEE Trans. Inf. Forensics Sec., 12, 1, 200–210, 2017. 15. Chen, Q. and Bridges, R.A., Automated behavioral analysis of malware: A case study of wannacry ransomware. Proc. IEEE 16th Int. Conf. Mach. Learn. Appl. (ICMLA), p. 454_460, Dec. 2017. 16. Ge, H., Yue, D., Xie, X.P., Deng, S., Hu, S.L., Analysis of cyber physical systems security issue via uncertainty approaches. Proc. Adv. Comput. Methods Life Syst. Model. Simulation, ch. 6, sec. 6, Springer, Nanjing, China, p. 421_431, 2017. 17. Tian, M., Wang, X., Dong, Z. et al., Injected attack strategy for false data based on lagrange multipliers method. Autom. Electr. Power Syst., 41, 11, 26–32, Jun. 2017. 18. Gong, J., Zang, X., Su, Q. et al., Survey of network security situation awareness. J. Software, 28, 4, 1010–1026, Nov. 2017. 19. Li, T., Su, S., Yang, H. et al., Attacks and cybersecurity defense in cyber-physical power system. Autom. Electr. Power Syst., 41, 22, 162–167, Nov. 2017. 20. Lee, H.-C., Jang, T.-I., Moon, K., Anticipating human errors from periodic big survey data in nuclear power plants. Proc. IEEE Int. Conf. Big Data (Big Data), pp. 4777–4778, Dec. 2017. 21. Feng, S. and Tesi, P., Resilient control under Denial-of-service: Robust design. Automatica, 79, 42_51, May 2017.

Cybersecurity and Privacy Fundamentals  377 22. Fritz, R. and Zhang, P., Modeling and detection of cyber attacks on discrete event systems, in: IFAC-PapersOnLine, vol. 51, p. 285_290, May 2018. 23. Zhu, W., Deng, M., Zhou, Q., An intrusion detection algorithm for wireless networks based on ASDL. IEEE/CAA J. Autom. Sin., 5, 1, 92_107, Jan. 2018. 24. Ding, D., Han, Q.-L., Xiang, Y., Ge, C., Zhang, X.-M., A survey on security control and attack detection for industrial cyber-physical systems. Neurocomputing, 275, 1674_1683, Jan. 2018. 25. Sun, Y.-C. and Yang, G.-H., Event-triggered resilient control for cyber physical systems under asynchronous DoS attacks. Inf. Sci., 465, 340_352, Oct. 2018. 26. Ding, D., Han, Q.-L., Wang, Z., Ge, X., A survey on model-based distributed control and filtering for industrial cyber-physical systems. IEEE Trans. Ind. Informat., 15, 5, 2483_2499, May 2019. 27. Peng, C., Sun, H., Yang, M., Wang, Y.-L., A survey on security communication and control for smart grids under malicious cyber attacks. IEEE Trans. Syst., Man, Cybern. Syst., 49, 8, 1554_1569, Aug. 2019. 28. Lima, P.M., Alves, M.V.S., Carvalho, L.K., Moreira, M.V., Security against communication network attacks of cyber-physical systems. J. Control, Autom. Electr. Syst., 30, 1, 125_135, Feb. 2019. 29. Tang, Y., Li, M., Wang, Q. et al., A review on research of cyber-attacks and defense in cyber physical power systems part two detection and protection. Autom. Electr. Power Syst., 43, 10, 1–9, May 2019. 30. Liu, T., Tian, J., Wang, J. et al., Integrated security threats and defense of cyber-physical systems, in: Acta Automatica Sinica, vol. 45, no. 1, pp. 5–24, Jan. 2019. 31. Evans, M., He, Y., Maglaras, L., Yevseyeva, I., Janicke, H., Evaluating information security core human error causes (IS-CHEC) technique in public sector and comparison with the private sector. Int. J. Med. Inform., 127, 109–119, Jun. 2019. 32. Vaidya, R., Cybersecurity breaches survey 2019, Department for Digital,Culture, Media and Sport63Cybersecurity Breaches Survey 2019: Statistical Release free of cost available under Open Government License, pp. 01–66, 2019. 33. Evans, M., Maglaras, L.A., He, Y., Janicke, H., Human behaviour as an aspect of cybersecurity assurance. Secur. Commun. Netw., 9, 17, 4667–4679, Nov. 2016. 34. Taniuchi, S., Aoyama, T., Asai, H., Koshijima, I., Training cybersecurity exercise facilitator: Behavior modeling based on human error. Proc. Int. Conf. Appl. Hum. Factors Ergonom., 7, 138–148, 2019. 35. Evans, M., He, Y., Maglaras, L., Yevseyeva, I., Janicke, H., Evaluatinginformation security core human error causes (IS-CHEC) technique inpublic sector and comparison with the private sector. Int. J. Med. Inform., 127, 109–119, Jun. 2019. 36. Mayer, P., Kunz, A., Volkamer, M., Reliable behavioural factors in theinformation security context. Proc. 12th Int. Conf. Availability, Rel. Secur. (ARES), pp. 1–10, 2017.

20 Changing the Conventional Banking System through Blockchain Khushboo Tripathi*, Neha Bhateja† and Ashish Dhillon Department of Computer Science & Engg., Amity University, Haryana, India

Abstract

The banking and financial service sector has understood the advantages of blockchain technology. This chapter explores the tremendous impact, implementation challenges, and ground-breaking potential of the blockchain technology in these organizations. The blockchain is a powerful technology that enables Bitcoin, Litecoin, Ethereum, and many more currencies to be open, anonymous, and cryptographically secured. The blockchain technology is capable of offering banks a huge reduction in billions of real money by providing alternate ways to manage banking systems. Banks being huge technology consumers in today’s world are always looking for opportunities to scale down the transactional costs including the amount of paper work and related processing they have to handle. Implementing blockchain would be a step in this direction by providing cost reduction as well reliability in transactions. It is self-executable, self-verifiable, and embedded into the blockchain that eliminates the need for trusted third-party systems, which helps in saving administration as well as servicing costs. The benefits of blockchain technology in financial sector are in terms of fraud reduction, smart contracts, trade finance, clearing, and settlements. This literature also provides an insight on the challenges related to security and computationally less exhaustive algorithms which are not explored much by the authors around the world and also finding a suitable solution for financial sector. Keywords:  Cryptography, exhaustive algorithms, fraud-reduction, trade finance

20.1 Introduction 20.1.1 Introduction to Blockchain A block chain is simply a ledger system used for financial recording keeping in form of transaction. The ledge itself and the transactions that it has are considered of high integrity as each of the transaction is digitally signed for maintaining the authenticity and immutability. The main feature of this system is that these digital entries are distributed to all the nodes deployed in the network. These additional participants act as members for providing a consensus about the ledger system at any given point as each and every node maintains its own copy of blockchain ledger. *Corresponding author: [email protected] † Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (379–404) © 2021 Scrivener Publishing LLC

379

380  The Smart Cyber Ecosystem for Sustainable Development “A Blockchain is a cryptographically secured decentralized Peer-to-Peer ledger system which uses consensus mechanism for validation of transactions.” A blockchain system starts with adding a new transaction. This transaction is broadcasted to all the nodes available in the network which then evaluates and verifies the transaction by implementing the algorithms know as consensus algorithm. In case of bitcoin blockchain, consensus mechanism is known as proof of work (PoW). In consensus, if majority of nodes verify the history and signature of the transaction, then it will be accepted and sent into a block to be added in the chain. If transactions fails to gain majority of nodes, then the transaction is denied and will not be added to the chain. This happens mostly when the value in the transaction does not match with the input and output of the transaction. Hence, transaction will be invalid [1]. The following are the key characteristics of the block chain system, which gives it an edge over conventional information server and ledger systems. • Peer-To-Peer: All participating nodes are equal and have equal authorities. There is no central authority and this allows all the nodes to transact without third parties. • Decentralized: Since there is no central authority, so to validate the transactions, all the nodes participate in verification through a consensus mechanism protocol in case of bitcoin blockchain which is known to be PoW. • Cryptographically Secured: Security is achieved in the system by deploying various cryptography techniques. • Include-Only: Information must be added the blockchain with time-­ consecutive request. This property recommends that, once data is added to the blockchain, it is basically hard to change that data and can be considered in every practical sense perpetual. • Consensus: This is the most difficult of all. This empowers blockchain to invigorate the record by methods for accord. This is what gives it the power of decentralization. No central authority is responsible for invigorating the record. Or maybe, any update made to the blockchain is endorsed contrary to demanding standards described by the blockchain that showed and added to the blockchain essentially after an understanding has been reached among each and every taking an intrigued peer/centers on the framework in Figure 20.1. With the high pace of the net exchanges, the Bitcoin cannot technique the entire exchanges at a fast rate. The blockchain innovation goes up against with the amplified scope of exchanges and requires the necessity for breakdown the inertness because of the huge size of squares. In spite of the fact that the predominant blockchain based on the most part security frameworks gives an elevated level of security, it normally manages the quantifiability issue. Therefore, mining the chain of squares needs higher technique assets as far as voltage and methodology power. The permanent idea of the blockchain innovation entangles the information get to even once there is a necessity for dynamic hang-on records illicitly, for example, in instruction frameworks [2]. The blockchain-based bitcoin devours higher vitality though giving security contrasted with antiquated cash exchanges. Additionally, there is a

Changing Conventional Banking System  381 IMMUTABLE Information Received

2

Transaction Initiated

3

Transaction Executed

4

Value Transfer

5

Settlement of Digital Assets

SMART CONTRACT

1

Tr an In sac iti ti at on or

SELF-ENFORCING

First Party

SELF-VERIFIABLE

n tio sac or n a Tr itiat In

Second Party

SELF-EXECUTABLE

Figure 20.1  Immutability in blockchain.

fundamental might want for breakdown the deficiency of simple interfaces and furthermore the absence of durable framework though implementation.

20.1.2 Classification of Blockchains Blockchains can be classified into three parts which do not contain databases or ledger technology and are often misinterpreted as blockchains. The following are the three types of blockchains: • Public blockchains (examples, Bitcoin and Ethereum) • Private blockchains (examples, Hyperledger and R3 Corda) • Hybrid blockchains (example, Dragonchain)

20.1.2.1 Public Blockchain These blockchains allow anyone to participate as they are open sources. Participants can be users, miners, developers, or community members. Transactions that are carried out in this section are happen to be fully transparent, which means that anyone can examine the details of the transaction. These blockchains are created so that they are totally decentralized, with no I substance or an individual controlling the transactions that are recorded inside the blockchain or the while they are been prepared. Open blockchains are frequently exceptionally control safe, as anybody is neighborly to join the system, regardless of where they have a place with (area or nationality). Thus, it turns amazingly difficult for specialists to close them down [3, 4]. In conclusion, open blockchains contain flag identified with them that is commonly intended to boost and prize members inside the system.

382  The Smart Cyber Ecosystem for Sustainable Development

20.1.2.2 Private Blockchain Another kind of chains is private blockchains, additionally alluded to as permissioned blockchains, and has assortment of striking contrasts from open blockchains. • Participants need consent to join the frameworks. • Transactions are private and are only open to condition individuals that are offered agree to hitch the framework. • Private blockchains are more united than open blockchains. These are noteworthy for endeavors who wish to collaborate and communicate data, yet need not bother with their delicate market information clear on an open blockchain. Such chains, by their tendency, are progressively brought together; the instances running the chain have imperative order over individuals and organization structures. Private blockchains may or likely would not have a token incorporated the chain.

20.1.2.3 Hybrid Blockchain Dragonchain is carried out in a singular place inside the blockchain biological system where it is a half-breed blockchain. It indicates the protection benefits of a permissioned and private blockchain with the security and straightforwardness advantages of an open blockchain. That gives organizations critical adaptability to select information that they have to frame open and straightforward and what information they have to remain private. The half-breed nature of Dragonchain blockchain stage is framed conceivable by one’s licensed interchain ability; this permits people to just interface with other blockchain conventions. Keeping in consideration a multi-chain system of blockchains, this usefulness makes it basic for organizations to work with the straightforwardness that they are attempting to discover, without yielding security and protection. Likewise, being able to post to different open blockchains straightforwardly builds the security of exchanges, as they appreciate the joined hash power being applied to the overall population chains.

20.1.2.4 Consortium Blockchain Consortium blockchains are, in some cases, considered a different assignment from private blockchains. The principle contrast between them is that consortium blockchains are administered by a gaggle rather than one substance. This methodology has one of the comparable advantages of an individual blockchain and will be viewed as a sub-classification of individual blockchains, as against a different kind of chain. This cooperative model offers some of the least difficult use cases for the benefits of blockchain, uniting a gaggle of “reticent foes”—organizations that cooperate yet additionally go up against each other. They can be progressively effective, both exclusively and aggregately, by cooperating on specific pieces of their business. Members in consortium blockchains could incorporate anybody from national banks to governments, to give chains.

Changing Conventional Banking System  383

20.1.3 Need for Blockchain Technology There is a spread of blockchain use cases and advantages to blockchain usage, the preeminent notable being esteem moves over the Bitcoin convention. For cryptographic forms of money like Bitcoin, blockchain takes care of an extremely explicit issue that had hampered past endeavors at building up an advanced cash. That issue is comprehended in light of the fact that the “twofold spend” wonder. We, as a whole, comprehend that the standard route during which we share things inside the advanced world is to make an imitation of what we have, like a pdf or picture, and sending that to an alternate individual. As you will envision, if this pdf was a dollar, both the sender and beneficiary would have indistinguishable duplicates of this dollar and possibly could both spend it. Blockchain innovation understood this by guaranteeing the beneficiary realizes that lone they need the dollar and accordingly the sender realizes that they not have it. Anyone who attempts to spend the dollar realizes that lone resulting beneficiary currently has the dollar.

20.1.3.1 Bitcoin vs. Mastercard Transactions: A Summary Most people have a Mastercard that they will use to buy things. But some even have Bitcoin at their disposal. In order to maximize their assets, people should know how to use them.

20.1.3.1.1 Outcomes • Bitcoin transactions work in peer-to-peer mechanism: just like a cash exchanged without an intermediary (person to person). • Bitcoin is still in early development. Hence, it is not widely accepted and used for exchanges only. • Credit card companies have been developing from last six years and hence are trusted but charges high fee. • Banking systems offers fraud protection, whereas bitcoin has no central authority except being cryptographically more secured.

20.1.3.1.1.1 Bitcoin Transactions

Satoshi Nakamoto, the designer of bitcoin, titled his unique white book on the point “A Peer-to-Peer Electronic Cash System.” This portrayal addresses the center contrasts among bitcoin and Mastercard exchanges. Bitcoin installments are practically equivalent to wire moves or money exchanges, where installment is “pushed” straightforwardly from one gathering to an alternate, without perusing another budgetary association. Installment preparing is executed through an individual system of PCs, and each exchange is recorded during a blockchain, which is open. Bitcoin is predicated on distributed innovation and depends on the blockchain and in this way the cryptography making sure about it, with none outsider oversight. When making a bitcoin exchange, it is not important to gracefully close to home recognizable proof, similar to your name and address.

384  The Smart Cyber Ecosystem for Sustainable Development

20.1.3.1.1.2 Credit Card Transactions

By far, Mastercard allows the customers to effectively transact and authorize because of its long development in financial activities. Mastercard allows vendor to pull a payment and pass it through various third parties to complete the transaction to the recipient account. For instance, a regular Visa exchange includes four steps: the dealer, the recipient (the monetary association that grants installments to the vendor), the issuers (bank of the cardholder), and hence the individual cardholder.

20.1.3.1.1.3 Key Differences

Bitcoin transactions are made using digitally generated addresses that varies with every transaction and a private key. Transactions can be made on cellular devices using features like QR codes. In a bitcoin transaction, no personal information travels through the network. Hence, it improves security using anonymous alphanumeric addresses. Whereas, debit/credit cards are storing data physically in the electronic chips of the card and can be easily saved using a card reader and can also be pulled from the cloud database. Since we comprehend what a blockchain is and thusly the contrasting kinds of blockchains how it has evolved in the technology. There is a spread of blockchain use cases and advantages to blockchain execution, principal notable being esteem moves over the Bitcoin convention. For cryptographic forms of money like Bitcoin, blockchain takes care of an extremely explicit issue that had hampered past endeavors at building up a computerized cash. That issue is comprehended in light of the fact that the “twofold spend” wonder. We, as a whole, comprehend that the standard route during which we share things inside the computerized world is to make a copy of what we have, like a pdf or picture, and sending that to an alternate individual. As you will envision, if this pdf were a dollar, both the sender and beneficiary would have indistinguishable duplicates of money and possibly happened. The innovation understood that guaranteeing the beneficiary realizes that solitary they need the dollar and consequently the sender realizes that they not have it. Any individual who attempts to spend the dollar realizes that solitary ensuing beneficiary currently has the dollar.

20.1.4 Comparison of Blockchain and Cryptocurrency Individuals are confounded on the contrasts among blockchain and digital currency. A relatable method of encircling this relationship is to coordinate it to an application on your telephone (for example, Uber or Whatsapp) and, in this manner, the stage on which that application is running (IOS or Android). Blockchain is that the stage and digital money is an application that sudden spikes in demand for the blockchain stage. The disarray stems incompletely from the very reality that the stage (blockchain) and digital money (Bitcoin) propelled at an identical time.

20.1.4.1 Distributed Ledger Technology (DLT) A blockchain might be a database; however, it varies from a standard database in that the information put away subsequently is not unified in one area. Rather, a record of the record is held by the entirety of the members inside the chain which will confirm the provenance of the entirety of the information that is entered. Think about it as a database without a

Changing Conventional Banking System  385 manager. This proposes that members do not have to accept any single individual or element with respect to the veracity of information. Appropriated Ledger Technology, or DLT, might be a class of database innovation that has blockchain innovation or qualities of a blockchain. Be that as it may, only one out of every odd blockchain might be a conveyed record. Inside the Dragonchain, there is no single blockchain. Every business hub and each blockchain application has its own blockchain, which will cooperate with the other blockchain or framework in the event that you need, utilizing interchain tech. On Dragonchain, there is likewise no requirement for verification of work, or confirmation of stake, practically like dispersed record innovation. The terms are now and then utilized equivalently; however, there are different sorts of appropriated records that are organized uniquely in contrast to blockchains. A few examples of these incorporate IOTA and Hashgraph, which are all the more precisely portrayed as DAGs or Directed Acyclic Graphs. While blockchain was the essential circulated record innovation (DLT), it is not the sole kind of DLT that one can consider. Blockchain is basically one kind of conveyed record. A blockchain might be a grouping of squares, and dispersed records do not require such a succession. Conveyed records need not bother with verification of work or confirmation of stake and offer—hypothetically—better scaling alternatives contrasted with blockchains like Bitcoin and Ethereum.

20.1.5 Types of Consensus Mechanism In the previous scarcely any years, we heard tons about how blockchains are getting the chance to change the business world and monetary exchanges. Not many of the preeminent pivotal parts of blockchain are speed, applications, and security. In our past blog entries, we have featured blockchain innovation’s potential for speed and applications. Yet, to work on an overall scale, an open record needs productive security [5]. Also, along these lines the security is because of the agreement calculation. Those that are familiar with the blockchain innovation comprehend that we keep the record exchanges synchronized over the system to ensure that records possibly endorsed when the members affirm exchanges; see Figure 20.2.

20.1.5.1 Consensus Algorithm: A Quick Background A consensus algorithm resembles Bitcoin’s PoW, which needs diggers to disentangle complex cryptographic numerical riddles that they get compensated with explicit measure of Bitcoins. It is essential to realize that each square which is added to the system must observe a gathering of accord rules. For example, Bitcoin’s accord rules are no twofold spending, right configuration of hinders, a specific measure of remuneration for diggers, and so on. Hinders that neglect to adhere to these accord rules will be dismissed. A blend of PoW accord calculations and, in this way, the agreement rules offers a strong and solid system which is secure and guarantees that every hub inside the system concede to a day-by-day worldwide condition of the blockchain. A consensus algorithm has three essential highlights upheld where its proficiency is regularly decided in Figure 20.3.

386  The Smart Cyber Ecosystem for Sustainable Development Consensus Algorithms

Scalability Scalable Non-scalable

Communication model Synchronous system

Category

Asynchronous system

Partial synchronous system

Failure modes

Random

Crash failure

Software failure

Deterministic

Transient failure

Byzantine failure

Monte Carlo

Omission failure

Temporal failure

Security failure

Enviromental pertubations

Leader free Leader based Las Vegas

Hybrid solution

Figure 20.2  Types of consensus mechanism. Security A consensus protocol is defined to be safe if all the nodes offer same output and the outputs are valid as per the rules. This is also called as consistency of the shared state.

Real-time Value A consensus protocol ensures real-time liveness, if all correct nodes participating in concensus eventually produce a value.

Fault Tolerance

01

02

03

A consensus protocol offers fault tolerance, if it can recover from a failure of a node participating in consensus.

Figure 20.3  Basic features of consensus protocol.

Without a focal delegate, the system taking an interest client that build up this procedure got the opportunity to concede to the legitimacy of what’s being added to the record (utilizing a gathering of predefined rules). An agreement must be gone after the greater part of the hubs inside the system. Yet, exactly how powerful it is to execute such an agreement stays a piece progressing till today. The sole objective of the agreement convention is to allow the hub to talk among themselves and to offer a standard arrangement of the approved exchange which might be added to the record. This is frequently planned to stop unethical miners from adding false transactions and blocks. The sort of mechanism to be used depends on the sort of network. Let us discuss few of the mechanisms.

20.1.6 Proof of Work As we have already discussed in our previous blogs, PoW is nowadays commonly used and is most robust among all other existing algorithms for blockchain platforms. Every miner

Changing Conventional Banking System  387

Proof-of-Work

Proof-of-Stake

Figure 20.4  Commonly known consensus algorithms.

node has to solve a cryptographical puzzle to find a new block which is then added to the chain after approval by other nodes in the network. In the wake of understanding the riddle, the appropriate response is then sent to different excavators and confirmed by them before being acknowledged to their individual duplicates of the record. The PoW strategy characterizes that the hubs must receive the fork which conveys work, and it is outlandish that the two contending forks will create ensuing square together; see Figure 20.4. Bitcoin network also takes care of double-spend problem as each node participates in consensus for verifying transaction utilizing PoW mechanism.

20.1.7 Proof of Stake Proof of Stake (PoS) might be a substitute methodology for PoW which needs less CPU calculations for mining. In spite of the fact that this is regularly likewise a calculation, and along these lines, the reason for existing is same as PoW, the technique is kind of various here. As just if there should arise an occurrence of PoW, a digger is remunerated by settling numerical issues and making new squares, in PoS, the maker of a substitution square is picked during a deterministic way, relying on its riches, additionally characterized as stake. This recommends inside the PoS instrument, and there is no square prize. In this way, the excavators take the exchange expenses. PoS component has its own advantages and disadvantages, and consequently, the real usage is very mind boggling; see Figure 20.5.

20.1.7.1 Delegated Proof of Stake Delegated Proof of Stake (DPoS) is incredibly not the same as PoS. Here, token holders do not chip away at the legitimacy of the squares without anyone else; however, they select agents to attempt to the approval for them. During a DPoS framework, there are as a rule in the middle of 21–100 chose delegates. The picked delegates are being changed intermittently and appointed a request to convey their squares. On the off chance that you have less number of representatives, it permits them to orchestrate themselves proficiently and make structured schedule openings to distribute squares. In the event that the agents miss their squares every day or distribute invalid exchanges, the token holders vote them out and

388  The Smart Cyber Ecosystem for Sustainable Development

Proof of Work

Proof of Stake

Requires complex mathe-

Requires coin holders chosen

matical calculations which

in a deterministic way that is

is called Mining.

called Staking.

Figure 20.5  Comparison of proof of work and proof of stake.

supplant them with another chose delegate. In contrast to PoW and PoS, in DPoS, excavators can work together to create squares. With a community oriented exertion and an incompletely incorporated procedure, DPoS has been prepared to run significant degrees which are quicker than the different agreement calculations.

20.1.7.2 Byzantine Fault Tolerance Byzantine Fault Tolerance (BFT) name originated from a response to “Byzantine general problem”, a sensible issue that specialists Leslie Lamport, Robert Shostak, and Marshall Pease clarified in an instructional exercise paper. BFT is becoming accustomed to fix the trouble of a maverick or inconsistent hub. On the off chance that any individual from the network sends conflicting data to others about exchanges, the dependability of the blockchain separates, and there is no focal position which will step in to address it. To disentangle this, PoW as of now offers BFT through its preparing power. On the contrary hand, PoS needs an increasingly clear arrangement. Hubs will routinely pick request to spot truth exchange. Utilizing a rendition of PoS which works with BFT appears the first encouraging way to deal with endorsing exchanges inside the blockchain.

20.2 Literature Survey 20.2.1 The History of Blockchain Technology Blockchain innovation must be one among the most significant advancements of the 21st century given the expanding influence it has on different areas, from money related to assembling, additionally, as instruction. Obscure to a few is that blockchain history goes back to the main 1990s [6–8]. Since its prevalence began growing 2 or 3 years back, assortment of utilizations has sprung up about underlining the sort of effect and it is bound to have in light of the fact that the race for computerized economies heats up as in Figure 20.6.

Changing Conventional Banking System  389

ORIGIN 19912008

2009

TRANSACTIONS 2010

2011

2012

2013

CONTRACTS 2014

2015

2016

APPLICATIONS 2017

2018

Figure 20.6  History of blockchain technology—Timeline infographic.

20.2.2 Early Years of Blockchain Technology: 1991–2008 Stuart Haber and W. Scott Stornetta imagined what a significant number of us have come to comprehend as blockchain, in 1991. Their first work included performing on cryptographic form that made sure about chain of squares, whereby no one could mess with timestamps of archives [9, 10]. In 1992, they redesigned their framework to incorporate Merkle trees that improved effectiveness in this manner, empowering the social occasion of more archives on one square. Nonetheless, it is in 2008 that blockchain gistory begins to acknowledge pertinence, due to the work one individual or gathering by the name Satoshi Nakamoto. Satoshi Nakamoto is authorize in light of the fact that the minds behind blockchain innovation. Practically, no one is comprehended about Nakamoto as individuals except him may be an individual or a gaggle of people that chipped away at Bitcoin, the essential utilization of the computerized record innovation. Nakamoto conceptualized the essential blockchain in 2008 from where the innovation has advanced and found its way into numerous applications past cryptographic forms of money. Satoshi Nakamoto discharged the essential white paper about the innovation in 2009. Inside the white paper, he gave subtleties of how the innovation was well prepared to fortify advanced trust given the decentralization angle that implied no one could ever be in control of anything. Since the time Satoshi Nakamoto left the scene and gave over Bitcoin improvement to other center engineers, the advanced record innovation has developed prompting new applications that structure the blockchain History.

20.2.2.1 Evolution of Blockchain: Phase 1—Transactions 20.2.2.1.1 Blockchain Version 1.0: 2008-2013: Bitcoin Emergence

To most of the population, blockchain and bitcoin are synonyms and therefore think that both are same, but in reality, blockchain itself is an independent technology which supports a lot of applications out of which cryptocurrency is one of the use cases.

390  The Smart Cyber Ecosystem for Sustainable Development Bitcoin was introduced to the world in 2008 through a white paper published by Satoshi Nakamoto titled it as “Peer-to-Peer Electronic System”. Satoshi nakamoto mined the first block known as genesis block which act as starting point in a chain and upcoming blocks will be added to it forming chain. Each block comprises of the hash of previous block to interconnect. Since the bitcoin came into existence, the industry is developing new ways to take advantage of the blockchains principles and capabilities. There are lot technologies which have inherited from this digital ledger system.

20.2.2.2 Evolution of Blockchain: Phase 2—Contracts 20.2.2.2.1 Blockchain Version 2.0:2013-15 Ethereum Development

In this present reality where advancement is the thing to address, Vitalik Buterin is among the developing rundown of engineers who felt Bitcoin had not yet reached there, when it came to utilizing the total abilities of blockchain innovation, together with the essential supporters of Bitcoin codebase. Worried by Bitcoin’s restrictions, Buterin began performing on what he felt would be a pliant blockchain which will perform in different capacities, furthermore to be a shared system. Ethereum was brought into the world out as a supplanting open blockchain in 2013 with added functionalities contrasted with Bitcoin, an improvement that has dressed to be a significant crossroads in Blockchain history. Vitalik buterin as the inventor of Ethereum made sure that it is not similar to bitcoin blockchain for that he enables Ethereum to record some extra assets such as slogans and quotes with the help of smart contracts. This feature made Ethereum more suitable for developing decentralized application. Authoritatively launched in 2015, Ethereum blockchain has developed to get one among the most significant uses of blockchain innovation given its capacity to help brilliant agreements wont to perform different capacities. Ethereum blockchain stage has likewise prevailing with regard to social affair, and an enthusiastic designer network that has seen it builds up a genuine biological system. Ethereum blockchain forms the principal number of every day exchanges on account of its capacity to help shrewd agreements and decentralized applications. Its market top has likewise expanded altogether inside the digital money space [11, 12].

20.2.2.3 Evolution of Blockchain: Phase 3—Applications 20.2.2.3.1 Blockchain Version 3.0:2018: The Future Scope

Blockchain history and development does not stop with Ethereum and Bitcoin. As of late, assortment of tasks has sprung up all utilizing blockchain innovation capacities. New tasks have looked to manage some of the lacks of Bitcoin and Ethereum also to emerge with new highlights utilizing blockchain abilities. A portion of the new blockchain applications incorporates NEO, charged in light of the fact that the main open source, decentralized, and blockchain stage propelled in China. But the nation has restricted digital forms of money, and it stays dynamic when it includes blockchain developments. NEO throws itself in light of the fact that the Chinese Ethereum having just gotten the support of Alibaba CEO Jack Ma since it plots to have an identical effect as Baidu inside the nation.

Changing Conventional Banking System  391

20.2.3 Literature Review The Table 20.1 below represents the survey of different literature that has already been published. Table 20.1  Survey of literature review. Title/Authors

Remarks/Outcomes

1. Blockchain Technology A Literature Survey by Ibrar Ahmed1, Shilpi2, Mohammad Amjad. Published in the year 2018.

• Talks about pure blockchain and its use cases. • Entry-level paper outlining the working, challenges, and application. • Impure and hybrid blockchain are not discussed

2. Bitcoin: A Peer-to-Peer Electronic Cash System by Satoshi Nakamoto. Published in year 2008.

• First ever paper on blockchain • Only proposed for payment systems • Addressed storage issue

3. Opportunities and Risks of Blockchain Technologies In Payments—A Research Agenda by Juho Lindman, Matti Rossi, and Virpi Kristiina Tuunainen. Published in year 2017.

• Issues related to competitive environment • Problem in integration with other platforms • Pricing strategy/fees in blockchain

4. A Survey on Consensus Mechanisms and Mining Strategy Management in Blockchain Networks by Wenbo Wang, Dinh Thai Hoang, Peizhao Hu, Zehui Xiong, Ping Wang, and Yonggang Wen Dong In Kim. Published in the year 2019.

• Influence of consensus algorithms from different perspectives • Review of BFT-based protocols • Reward/incentive compatibility in protocol design • Strategy adoption by consensus participants

5. Current State of Blockchain Technology A Literature Review by Soto Méndez Jomar. Published in year 2018.

• Potential uses in government • Great opportunities for private sector • Breakthrough in accounting bringing transparency

6. The Blockchain Technology and Its Applications in the Financial Sector by Laura Jutila. Published in year 2017.

• Bitcoin vs. blockchain • Smart contracts • Security-based trading

7. In Search of an Understandable Consensus Algorithm by Diego Ongaro and John Ousterhout. Published in year 2014.

• Limitations of various algorithms • Factors for effective algorithm

8. Blockchain By Example by Bellaj Badr, Richard Horrocks, Xun (Brian) Wu. Published in year 2018.

• Understanding by Building a blockchain • Dapps and environments

9. A Survey of Blockchain from the Perspectives of Applications, Challenges and Opportunities by Ahmed Afif Monrat, Olov Schelen, and Karl Andersson. Published in year 2019.

• Trade-off of the technology

392  The Smart Cyber Ecosystem for Sustainable Development

20.2.4 Analysis The Table 20.2 below represents the comparison of different consensus algorithms with respect to Genetic Parameters. The Table 20.3 below represents the comparison of different consensus algorithms with respect to the performance related parameters. Table 20.2  Comparison of the consensus algorithms with respect to generic parameters. Parameters Consensus algorithms

Date

Blockchain type

Consensus category

Consensus finality

PoW

2008

Permission-less

Based on computational power (15,546, 745,765,529)

Proof based

Probabilistic

Elastico

2016

Permission-less

Based on computational power

Proof based

Absolute/instant irreversibility

PoV

2017

consortium

Vote-based mining

Vote based

Probabilistic

Mining difficulty

Table 20.3  Comparison of the consensus algorithms with respect to performance-related parameters. Parameters Energy/Resource consumption

Adversary tolerance model

Consensus algorithms

Transactions per day

Scalability

Latency

PoW

344,615

Not scalable

10 minutes (approxim­ately depending on the difficulty)

538 KWh electricity consumption

1 Mb (approxim­ ately 2,303 transactions)

No node can possess more than 50% of the computational power.

Elastico

-

Linear

103–110 seconds with around 400–800 nodes

-

Constant, 5 Mb/node

Faulty process can have up to 1/4th of the computational capacity

PoV

-

Not scalable

0.25 minutes for transaction verification

Very less compared to pow

Variable

Faulty processes 50% of processes

Size of block

20.3 Methodology and Tools 20.3.1 Methodology The analysis has predicated on subjective research techniques, utilizing work area investigation, writing audit, meetings, and contextual investigations to get proof. With a rising innovation like blockchain, with practically every day industry declarations and posts on master media, the usage of subjective strategies at present speaks to a down-to-earth approach in connecting

Changing Conventional Banking System  393 with the point when investigation on the theme is at an early stage, and where contextual analyses including the blockchain and training are exploratory and/or pilot activities. Our research approach involves the following: • In detail study of aspects and characteristics of current consensus mechanisms, aspects involves nature of consensus mechanism, concept, and, therefore, the significance of mechanism within the system; security aspects of the consensus algorithm and vulnerability and cases of failure in algorithm; characteristics like delay, computational requirements, time efficiency, and adaptability in various financial models; a critical comparison of all the prevailing most ordinarily used consensus algorithms for analysis. • Development of other for existing algorithms; suggestions and fundamental improvements to be done on the idea of critical appraisal; testing the algorithms on the testing networks and network simulation tools using dApps. • Based on the results of testing and adaptableness of the consensus algorithms, suggestion for best algorithm for a specific quite application.

20.3.2 Flow Chart STEP 1 Standard Reference

STEP 2 Problem Analysis -Per Transaction -Average Block Generation Time

STEP 3 Nbitcoin Core - New Algorithm

STEP 4 Run

STEP 5 Analysis

Figure 20.7  Flow chart of the process.

Figure 20.7 depicts the flow chart of the process that has been implemented.

394  The Smart Cyber Ecosystem for Sustainable Development

20.3.3 Tools and Configuration Tools that are used for the experimentation are as follows: • • • • •

Testnet bitcoin core (bitcoin network simulator) Nbitcoin core (network simulator) Visual studio Blockchain explorer Sha256 hashing algorithms

20.4 Experiment 20.4.1 Steps of Implementation Step 1: To select the standard algorithm: Here, PoW is chosen the standard reference. To execute, this PoW Bitcoin core open source blockchain simulation tool is used.

Figure 20.8  Screenshot of bitcoin core.

Figure 20.8 depicts the Screenshot of bitcoin core. Step 2: Data calculation: Current transactions done per second are calculated Bitcoin blockchain transaction data.

Changing Conventional Banking System  395 Confirmed Transactions Per Day 310,253 452,624 401,491 350,357 299,224 248,091 2019-04-17

blockchain.com/charts

2020-04-15

Figure 20.9  Screenshot of transactions per day.

Figure 20.9 represents the Screenshot of transactions per day. Step 3: Issue within the data.



For the knowledge of the typical size of Mastercard transactions, it is about 500 bytes in average and its throughput is 5,000 transactions per second.

Median Confirmation Time 6.1 min 19.5 min 16.2 min 13.0 min 9.7 min 6.4 min 2020-03-17

blockchain.com/charts

2020-04-15

Figure 20.10  Median time confirmation of transactions.

Figure 20.10 represents the Median time confirmation of transactions. These figures translate to, in terms of blockchain (example, QWC), as follows: 5,000 txs per second × 60 seconds per minute × 2 minute per block. This translates to 5,000 × 60 × 2 = 600,000 transactions per block. The size of every block has got to be on the brink of 300 Mb per block (600,000 transactions per block × 500 bytes per transaction).

396  The Smart Cyber Ecosystem for Sustainable Development The current block size of QWC is 1 Mb, and it can delay to about 1,000 transactions per block. If you check out it this manner, the blockchain’s performance seems really slow. The more decentralized a blockchain is, the slower its performance are often. The highest TPS is, needless to say, provided by the payment networks like Visa and Mastercard. Mastercard maintains 99.999 the availability and may process 3.4 billion transactions per day at 38,000+ transactions per second, with a mean reaction time of 140 ms. Step 4: Modification of algorithm.

Figure 20.11  Algorithm modification.

Figure 20.11 shows the modification done in algorithm. Step 5: Execution of modified algorithm. All the logs are collected and analyzed for result.

Figure 20.12  Execution of new algorithm.

Figure 20.12 shows the screenshot of execution of new algorithm.

Changing Conventional Banking System  397

20.4.2 Screenshots of Experiment

Figure 20.13  Block services.

Figure 20.13 represents the Block services.

Figure 20.14  Block services.

Figure 20.14 represents the Block services.

398  The Smart Cyber Ecosystem for Sustainable Development

Figure 20.15  Transaction services.

Figure 20.15 represents the Transaction services.

Figure 20.16  Working directory.

Figure 20.16 represents the Working directory.

20.5 Results Program with the modified algorithm was executed for 5 hours to obtain substantial amount of data for analysis. Data of total 1,000 blocks was recorded with timestamp and hashes of each block. It took an average of 17 seconds to generate one block.

Changing Conventional Banking System  399 no. of transactions per day

myChain experiment

9148235

Current transaction system(master card, etc)

3400000000

Bitcoin Blockchain

310253 2E+09

3E+09

4E+09

no. of transactions per day

Current transaction system(master card, etc)

Bitcoin Blockchain

no. of transactions per day

310253

myChain experiment

3400000000

9148235

Figure 20.17  Number of transactions per day.

Figure 20.17 represents the number of transactions per day. Overall, Visa and Mastercard continue to have one of the fastest transaction speeds across several different payment networks. Keep in mind that cryptocurrency and blockchain technology is still in the very early stages. Visa was founded in 1958 and has had 60 years to improve and grow its payment network capabilities. Think of giving current cryptocurrencies like Bitcoin, Ethereum, Litecoin, and Stablecoin, where the amount of time of the current payment system has to evolve and the race can be easily won by them. For now, Bitcoin and other platforms are developing at a faster rate and better adoption [13–15]. energy consumption per year (Twh)

10

myChain experiment Current transaction… 28.67

Current transaction system(master card,etc)

Bitcoin Blockchain

myChain experiment 39

Bitcoin Blockchain energy consumption per year(Twh)

0

5

10 28.67

15

20

25

39

30

energy consumption per year (Twh)

Figure 20.18  Energy consumption per day.

Figure 20.18 represents the energy consumption per day.

35

40

45 10

400  The Smart Cyber Ecosystem for Sustainable Development block size (in MB)

myChain experiment

2

Current transaction system(mater card, etc)

300

Bitcoin Blockchain

2 0

50

100

150

200

250

300

350

block size (in MB)

Figure 20.19  Block size.

Figure 20.19 depicts the block size. But in terms of size of the blocks for Visa and Mastercard, large size blocks may create overhead for networks and could lead to congestion. In such cases, Merkle tree algorithm benefits in trimming down the data sizes for bank servers.

20.6 Conclusion • One of the key issues which are influencing the adoption of blockchain is that it can be run on low powered devices which are achieved by the light weight algorithm. • Reflection algorithm requires a 10 twh of energy which is a third of the current server systems that enables it fit for not only banks but also for small businesses • Improvement of the PoW algorithm is almost 30 times which is evident from the fact that it improves the number of transactions per day from 310,256 to 910,2562. • Blockchain is still in early development stages and has a lot of potential in form of smart contracts which makes it fit in almost every commercial sector which involves EMI or loans. • Digital record keeping which can be used as private and public both will be an added benefit for government transactions which needs to be private and the trust funds can be on public blockchain which brings transparency to the system. • Identification of the key issues which are impacting approach producers and other key partners in considering the utilization of blockchain innovation as a worth included recommendation inside the money related division scene. • Determined whether the technology can be used to fit-for-purpose for the recording of transactions within required time constraints, and the effectiveness of the consensus mechanism protocols. • Discussed the usage of blockchain technology to bridge the legitimate need for peer-to-peer transactions without involving the third party services.

Changing Conventional Banking System  401

20.7 Future Scope Organizations such as Walmart are hoping to use the immutability and shared possession or consortium highlights of blockchain to empower improved following and detectability of food items, prompting better sanitation in its stores. Banks are utilizing private blockchains to tokenize (digitize) their own inner resources, permitting them to move finances inside sparing numerous dollars in costs. Organizations like BitPesa are empowering organizations in locales with poor financial administrations to move supports all the more effectively across outskirts. Accounting firms are seeing the potential that straightforwardness and immutability offer their review and bookkeeping groups.

20.7.1 Blockchain as a Service (BaaS) is Gaining Adoption From Enterprises Early adopters of blockchain innovation, as other early innovation adopters, have had in contact the brunt of the difficulties with expanding on any new stage. Arrangement challenges, poor engineer apparatuses, and operational issues have for quite some time been the quality with standing up new usage. We have presently arrived at some degree, be that as it may, where BaaS can make fabricating a blockchain application progressively available to anybody. As characterized by Investopedia, BaaS is a contribution that grants clients to use cloud-based arrangements (like Dragonchain) to host, assemble, and utilize their own blockchain application and smart contracts. E-commerce Global Payments

Digital Rights

Escrow

ts rac nt

Dig ita lC

Remittance P2P Lending

cy ren ur

Sma rt C o

Wagers

Microfinance

THE BLOCKCHAIN Equity

Debt Crowdfunding Derivatives

in g ep

cu

r it

ie s

Title Records

Ke

Se

Private Markets

Healthcare

R

r eco

d

Ownership Voting Intellectual Property

Figure 20.20  Future scopes of blockchain.

Figure 20.20 represents the future scope of block chain technology.

402  The Smart Cyber Ecosystem for Sustainable Development

References 1. Peters, Gareth, W. and Panayi, E., Understanding modern banking ledgers through blockchain technologies: Future of transaction processing and smart contracts on the internet of money, in: Banking beyond banks and money, pp. 239–278, Springer, Cham, 2016. 2. Kogias, E.K. et al., Enhancing bitcoin security and performance with strong consistency via collective signing. 25th {USENIX} Security Symposium ({USENIX} Security 16), 2016. 3. Daguerre, C., María, J., Villa Perez, A., Systematic Literature Review of the use of Blockchain in Supply Chain, 2017. 4. Casino, F., Dasaklis, T.K., Patsakis, C., A systematic literature review of blockchain-based applications: current status, classification and open issues. Telematics Inf., 36, 55–81, 2018. 5. Shen, C. and Pena-Mora, F., Blockchain for Cities—A Systematic Literature Review. IEEE Access, 6, 76787–76819, 2018. 6. Wang, W., Hoang, D.T., Hu, P., Xiong, Z., Niyato, D., Wang, P., Wen, Y., Kim, D., I, A survey on consensus mechanisms and mining strategy management in blockchain networks. IEEE Access, 7, 22328–22370, 2019. 7. Monrat, A.A., Schelén, O., Andersson, K., A Survey of Blockchain From the Perspectives of Applications, Challenges, and Opportunities. IEEE Access, 7, 117134–117151, 2019. 8. Blockchain network simulator, 2016. [Online]. Available: https://dsg- titech.github.io/simblock/ 9. Testnet, Testing network 2017. [Online]. Available: https://bitcoin.org/en/glossary/testnet 10. Ahmed, I. and Shilpi, M.A., Blockchain Technology A Literature Survey. Int. Res. J. Eng. Technol., 5, 1490–1493, 2018. 11. Nakamoto, S., Bitcoin: A peer-to-peer electronic cash system. Manubot, 2019, [online] Available: https://git.dhimmel.com/bitcoin-whitepaper/. 12. Lindman, J., Tuunainen, V.K., Rossi, M., Opportunities and risks of Blockchain Technologies–a research agenda. Proc. 50th Hawaii Internat. Conf. Syst. Sci., 1533–1542, 2017. 13. Feller, W., An introduction to probability theory and its applications, (2nd ed.). John Wiley, 1957. 14. https://mlsdev.com/blog/156-how-to-build-your-own-blockchain-architecture 15. https://www.investopedia.com/articles/forex/042215/bitcoin-transactions-vs-creditcard-transactions.asp 16. https://cointelegraph.com/news/research-bitcoin-can-beat-visa-mastercard-to-top-worldpayment-system-in-10-years 17. Wang, W., Hoang, D.T., Hu, P., Xiong, Z., Niyato, D., Wang, P., Wen, Y., Kim, D., I, A survey on consensus mechanisms and mining strategy management in blockchain networks. IEEE Access, 7, 22328–22370, 2019. 18. Mendez, J.S., JomarCurrent State of Blockchain Technology A Literature Review, 2018. Available: https://www.academia.edu/37967532/Current_State_of_Blockchain_Technology_A_ Literature_Review 19. Jutila, L., The blockchain technology and its applications in the financial sector, 2017. Available: https://core.ac.uk/download/pdf/84757723.pdf 20. Ongaro, D. and Ousterhout, J., In search of an understandable consensus algorithm. In 2014 {USENIX} Annual Technical Conference ({USENIX}{ATC} 14), pp. 305–319, 2014. 21. Badr, B., Horrocks, R., Wu, Xun (Brian), Blockchain By Example, 2019. 22. Monrat, A.A., Schelén, O., Andersson, K., A survey of blockchain from the perspectives of applications, challenges, and opportunities. IEEE Access, 7, 117134–117151, 2019. 23. Wahab, A. and Mehmood, W., Survey of consensus protocols. arXiv preprint arXiv, 2, 1810.03357, 2018. 24. Dai, W., b-money, 1998. http://www.weidai.com/bmoney.txt.

Changing Conventional Banking System  403 25. Massias, H., Avila, X.S., Quisquater, J.-J., Design of a secure timestamping service with minimal trust requirements. 20th Symposium on Information Theory in the Benelux, May 1999. 26. Haber, S. and Stornetta, W.S., How to time-stamp a digital document. J. Cryptol., 3, 2, 99–111, 1991. 27. Bayer, D., Haber, S., Stornetta, W.S., Improving the efficiency and reliability of digital time-stamping. Sequences II: Methods in Communication, Security and Computer Science, pp. 329–334, 1993. 28. Haber, S. and Stornetta, W.S., Secure names for bit-strings. Proceedings of the 4th ACM Conference on Computer and Communications Security, pp. 28–35, April 1997. 29. Merkle, R.C., Protocols for public key cryptosystems. Proc. 1980 Symposium on Security and Privacy, IEEE Computer Society, pp. 122–133, April 1980.

21 A Secured Online Voting System by Using Blockchain as the Medium Leslie Mark1, Vasaki Ponnusamy1*, Arya Wicaksana2, Basilius Bias Christyono2 and Moeljono Widjaja2 Faculty of Information and Communication Technology, Universiti Tunku Abdul Rahman, Kampar, Malaysia 2 Department of Informatics, Universitas Multimedia Nusantara, Scientia Boulevard, Tangerang, Banten, Indonesia 1

Abstract

The secured online voting system aspires to achieve its objective by developing a website for eligible voters to cast their votes online in a secured manner. This system is designed to ease the difficulties and tedious procedures that voters would have to endure when they seek to cast their votes during the polling day. Secure computer algorithms would immensely improve efficiency while reducing the amount of time for the entire operation and eradicating human errors. When diversifying traditional voting systems into an online voting system, an essential factor that should be considered is cybersecurity. Online voting systems are vulnerable to attacks performed by hackers. This brought about the need to develop a fully secured online voting system. Various researches have been conducted using multiple algorithms to construct a secured online voting system that includes steganography, visual cryptography, hybrid crypto realm, encryption, digital signature, and biometric security. This paper reviews various secured online voting systems in the body of knowledge and utilizing public-key cryptography, digital signature, steganography, and many more. A particular emphasis would be given on blockchain technology’s role in the secured online voting system, mainly to preserve the voting tracking system’s integrity and authenticity. Keywords:  Blockchain, digital voting, secured voting system

21.1 Blockchain-Based Online Voting System 21.1.1 Introduction The use of a blockchain system was initially proposed for peer-to-peer payment systems. These systems are generally to facilitate cash transactions through the internet. The system also intends to do so without the need to rely on external financial institutions. Typically, blockchain systems are secured by design. Blockchain provides additional security layers *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (405–430) © 2021 Scrivener Publishing LLC

405

406  The Smart Cyber Ecosystem for Sustainable Development Block 1 Header

Block 2 Header

Block 3 Header

Hash of Previous Block Header

Hash of Previous Block Header

Hash of Previous Block Header

Merkle Root

Merkle Root

Merkle Root

Block 1 Transactions

Block 2 Transactions

Block 3 Transactions

Figure 21.1  Linking of multiple blocks in a blockchain system.

that neither steganography [1] nor cryptography [2] approaches could provide. Among the first systems that used this approach was Bitcoin. This system solely depends on cryptography to secure transactions that involve the use of Bitcoins. In short, blockchain is an ordered structure of data that contains multiple blocks of a transaction. Each block in the chain will have to be linked to the previous block in the chain for the system to be considered to be using this approach. The first block in the chain is the foundation of the stack. New blocks get layered over the previous ones to form a stack. As more blocks are attached to this foundation, the chain gets longer, creating the blockchain. A representation of blockchain systems is shown in Figure 21.1.

21.1.2 Structure of a Block in a Blockchain System

Block Size

Block Header Transaction Counter

Block Transaction

Figure 21.2  Typical block structure in a blockchain system.

21.1.3 Function of Segments in a Block of the Blockchain 1. Block Size: Determines the size of the whole block. This segment comprises of 4 bytes in size. 2. Block Header: Encrypts a unique hash. This is where the encryption occurs for each block in the blockchain sequence. This segment comprises of 80 bytes.

Secured Online Voting System Using Blockchain  407 3. Transaction Counter: Number of transactions/blocks that follow after this particular block. The segment size is between 1 and 9 bytes (depending on the number of blocks). 4. Transaction: Contains the transaction that is stored in this particular block. The size of this segment varies depending on the transaction size (Figure 21.2).

21.1.4 SHA-256 Hashing on the Blockchain In such systems, each block in the stack (blockchain) will be identified by a hash that would be placed on the block’s header. The hashing algorithm used is the Secure Hash Algorithm (SHA-256). This will be used to generate a near idiosyncratic fixed-size 256bit hash. It should be noted that this algorithm was designed by the National Security Agency (NSA) during the year 2001. The function of this hashing algorithm is basically to get a plaintext as an input. The input can be of any size. Upon obtaining the input, the algorithm encrypts it to a 256-byte binary value. This is strictly a one-way function. An example of the hashing algorithm is shown in Figure 21.3. Figure 21.4 shows how each block broken up and encrypted. The multi-layer hashing algorithm using SHA-256 ensures that the system is well secured regarding storing vote casting information by ensuring vote and voter integrity. As mentioned earlier, each header in each block in the blockchain will contain information that links the block to the previous block in the chain. These links to the previous blocks would eventually lead to the foundation (start of the blockchain). This would be the same concept used in the online voting system when its systems use a blockchain algorithmic approach. Assuming we have two candidates contesting in a constitution— Candidate A and Candidate B. Voter 1 accesses the system as a new voter and votes for

Hash

SHA-256 Hash Function

Plain text Input Sequence

acee16a0713f6433 5b53fd62c861d216 57c4329f57ae0c82 c9a5c0ee172b7237

Figure 21.3  SHA-256 Hash Function.

Top hash Hash 0

Hash 1

Hash 0-0

Hash 0-1

Hash 1-0

Hash 1-1

Data block 000

Data block 001

Data block 002

Data block 003

Figure 21.4  SHA-256 Hash Function after block segmentation.

408  The Smart Cyber Ecosystem for Sustainable Development Candidate A. Voter 1 becomes the foundation for the number of votes cast for Candidate A. Hence, a new block is created. Now, let us assume Voter 2 also votes for Candidate A. This block will be linked to the Voter 1 block in the blockchain. Every subsequent voter who casts their votes for Candidate A will be linked to the previous block included in the blockchain. The number of votes is cast for Candidate A. Voter 1 will remain the foundation for this blockchain. Each new voter (new block) will be encrypted using the HSA-256 encryption algorithm before it is inserted into the blockchain. Vote integrity is maintained when using this approach. Figure 21.5 shows how the system uses the approach in the system. In Figure 21.6, we can see how the blockchain approach is used in online voting systems. A candidate’s vote count block is the foundation for the blockchain. Every subsequent vote cast for the candidate will be linked to the block preceding it. This ensures all votes cast in the system is linked to a legitimate voter. They are thus securing the system from duplicate and phantom voters. User information is encrypted using the SHA256 hashing algorithm, ensuring the voters’ anonymity, which is among the system’s strengths. Vote integrity is ensured by the blockchain approach, where only legitimate accounts can vote. Hence, only they can create a new block that can be inserted into the blockchain. Figure 21.7 and Figure 21.8 shows the interaction and general framework of blockchain-based online voting system.

Information concerning the new voter

New Block HSA-256 Hash Function

Information concerning the previous voter

- Hash value - vote

Figure 21.5  Insertion/encryption of a new block into the blockchain.

Candidate 1

vote 1

vote 2

vote 3

vote n

Candidate 2

vote 1

vote 2

vote 3

vote n

Candidate n

vote 1

vote 2

vote 3

vote n

Figure 21.6  Vote count process in online voting system using blockchain approach.

Secured Online Voting System Using Blockchain  409

21.1.5 Interaction Involved in Blockchain-Based Online Voting System

Online Voting System

Registered Voter Database

Voter

PC (User Device) SHA-256 Encryption Algorithm

Registered Candidate Database

Blockchain

Figure 21.7  General interaction in online voting system using blockchain.

21.1.6 Online Voting System Using Blockchain – Framework

Server PC

Client PC

Online Voting System using Blockchain Web application

Voter

Request

Cleansing

Encrypting/ Signing of Vote

Vote Forwarding Server

Response

Candidate Database

Vote Decryption Tallying

Election Commission Employees and System Administration

Voter Database

Internet

PC (Web-Browser)

Election Authorities

Database

Votes Decryted For Tallying

Generating Results

Figure 21.8  Potential framework of an online voting system using blockchain.

410  The Smart Cyber Ecosystem for Sustainable Development

21.2 Literature Review 21.2.1 Literature Review Outline Many kinds of research have been conducted to tackle the security concerns that are very much associated with online voting systems. Different cryptographic algorithms have been used in the past in various ways, with each of them has its strengths and weaknesses. This portion of the report will dive in-depth into all related research papers that use different algorithms to develop a secured online voting system.

21.2.1.1 Online Voting System Based on Cryptographic and Stego-Cryptographic Model This research paper uses steganographic and cryptographic algorithms to develop a secured online voting system. Steganographic systems would consist of a secret message, cover data, and the stego message [3]. In this system, the secret message is the part of the data that contains sensitive information that is therefore needed to be hidden from potential attackers. The cover data in the algorithm serves as a container used to cover the secret message. This technique utilizes unused data in the covert data with bits containing information that is different and invisible. Unused data in the covert data mentioned above could refer to data that makes up sound, text, graphics, and videos. Extensive data that would make up good quality media such as video, image, and sounds usually contain redundant information. This is the part of the data that would be used to envelop the secret message. As shown in Figure 21.9, it explains how unused data acts as a cover for the secret message to pass through the network. Using a stego object in an online voting system can ensure the voter’s selected candidate remains confidential. This is essential for an online voting system as it protects voters from attackers who try to decipher or manipulate the casted vote. Upon obtaining the unused data, the data embedding procedure would commence. As shown in Figure 21.10, the data would be covered using the cover image during the coding phase. The key is generated using the cover image and the files intended to be hidden before moving across the transmission phase. Upon reaching the administrator’s end, the decoding phase would commence. The shared key would be used to decode the hidden files.

stego-key cover-object message to embed

embedding

stogo-key stegoobject

cover-object extracted message

extraction receiver

sender network

Figure 21.9  Steganographic system framework [3].

Secured Online Voting System Using Blockchain  411 Coding Phase

Decoding Phase

Cover Image (C) C M

gk(C,M)

Key(K)

Communication Channel Attack Transmission Phase

gk(C,M)

^ M Recovered File

Key(K)

Files to Hide (M)

Figure 21.10  Steganographic embedding process [3].

At the end of the procedure, the hidden file would then be recovered. The strengths of this system include the ability to encode secret messages using various forms of data cover and also the ability for the system to camouflage hidden data as it would appear the same as the input data, thus making it difficult to be identified.

21.2.1.2 Online Voting System Based on Visual Cryptography In this particular research paper, a mathematical algorithm is designed to decrypt the message without one besides the intended receiver. This system achieves this by generating visual cryptography shares using the basic cryptographic model [4]. Each generated share is then encrypted using the RSA algorithm of the public key cryptography. During this phase, the shares produced have to be well protected. This is because attackers may try to form new shares that do not coincide with the existing ones or alter the bit arrangement or sequence, which would cause the data decryption at the receiver’s end to be incorrect. This is shown clearly in Figure 21.11. The secret image is broken down and forms shares. This takes place in the first phase. These shares are each encrypted using RSA encryption in the second phase. After that, these encrypted shares travel through the network until it reaches the recipient. Here, the shares are decrypted using RSA, and that comprises the third phase. Finally, the image shares are merged back to form the secret image at the receiver’s end, completing the fourth phase. This proposed system is well secured from attackers who wish to decrypt the information traveling in the network. Nevertheless, this system has potential weaknesses as attackers may significantly affect the system’s content of shares. Figure 21.12 shows the pixel encoding schemes of the proposed system. It is quite apparent that there are multiple ways image shares can be generated to five out of the similar output. This algorithm can be further enhanced by increasing the number of bits used to represent a secret image share, thus making it much harder to decrypt when in transmission in a network. That is among the strengths of the system. However, a noticeable weakness is that the contrast of the reconstructed image would not be maintained, and potential attackers may add new shares or modify existing ones that would affect the outcome of the image at the receiver’s end.

412  The Smart Cyber Ecosystem for Sustainable Development Secret Image

Share 1

Share 2

Phase 1

Encryption using RSA

Encryption using RSA

Phase 2

Decryption using RSA

Decryption using RSA

Decoded Secret Image

Phase 3

Phase 4

Figure 21.11  Proposed system framework using visual cryptography [4].

Figure 21.12  Sample image share generation in proposed system [4].

21.2.1.3 Online Voting System Using Biometric Security and Steganography This research paper uses a combination of biometric security elements and steganography [5]. Biometric identity is used in this system for added security purposes. Steganography is used in this system as it serves as a media cover. The redundant bits of this media coverage would be used to cover the secret message. Insertion into the least significant bit (LSB) is the standard practice. If the image comprises more bits, then more bits can be used in each pixel of the image to cover the secret message. However, the downside of using the algorithm is that if these modifications were inserted into the LSB, it would be detectable by eavesdroppers as there would be apparent distortions that would be visible in the end image when examined

Secured Online Voting System Using Blockchain  413 closely. This system also requires users to use their personal, National Identification (NID) numbers. This value is used to distinguish different voters and serve as an attribute to store voter information in the database. Using NID elements, every eligible voter’s biometric characteristics would be made available in the system. Therefore, all biometric information that would be found in the database would be unique and distinguishable. This will serve as a crucial element to keep track of voter turn up and duplication of votes during the commencement of a general election. As the voter logs in into their account, Optical Character Recognition is used for identification purposes at the server-side. Once validated, the voter will be allowed to cast their votes. As shown in Figure 21.13, it describes clearly how this online voting system uses steganography to form a stego image to transmit a secret message through a network. Figure 21.14 shows instances of how a cover mage is created. The redundant bits in this image would then be used as a cover to envelop the secret message. Algorithms such as replacing the LSB or the most significant bit (MSB) to contain the secret message can be used. LSB is an easier way to store bits into a pixel; however, it is less secured as image distortions can be detected easily by attackers. Using the MSB is harder to store bits into the image but is more secure as image distortions are less detectable. The secret key that also plays a pivotal role is the operational proceedings in this online voting system. The use of biometric measures in ensuring that the online voting system is fully secured gives a significant advantage in terms of this system’s security strengths. The drawback of this system, however, would be the expansion of the secret key. As mentioned before, this is a crucial element in this system and must be kept in full secrecy. However, since this an online voting system, there would be many users using this system. A short private key has to

Figure 21.13  Creation of the cover image [5].

Cover Image (256x256px)

Secret message

Stage Image (256x256px)

F–1

F

Key Image (256x256px)

Key Image (256x256px)

Stage Image (256x256px)

Figure 21.14  Use of stego image to envelope secret message [5].

Secret message

414  The Smart Cyber Ecosystem for Sustainable Development be used in order for users to remember them. A short private key would mean that it would be easier to decipher the private key using attacking algorithms such as a brute force attack.

21.2.1.4 Cloud-Based Secured Online Voting System Using Homomorphic Encryption This research paper’s proposed online voting system attempts to utilize Paillier’s encryption technique [6]. Homomorphic encryption can be very beneficial as operations that use the stored data to identify the polling winner without needing to decrypt the data in question. Phases in using this online voting system include the addition of candidate’s phase where candidates of a particular constituency have also added the position they are vying for, as shown in Figure 21.15. Next are the voter registration and validation phase, shown more clearly in Figure 21.16. In this phase, the voter would need to register themselves and validate the particular account in question to be eligible to cast their votes during the polling Web Server DB Candidate List Candidate ID

Name

Qualification

Contesting Position

1 2 .. m

A B .. ..

--.. ..

--.. ..

Admin adds candidate list with their bio-data and position they are contesting for from existing offline data Admin

Offline Candidate List

Web Browser

Figure 21.15  Candidate insertion into database [6].

Web Server DB Voter List Voter ID

Name

Contact Info Address

1 2 .. ..

X Y .. ..

--.. ..

Password

--.. ..

Offline Voter List

Voter Registration Form Android App

Voter Registration Form Web Browser Share Voter Login Password and Public Key

Figure 21.16  Eligible voter registration and validation [6].

Admin Web Browser

Secured Online Voting System Using Blockchain  415 day of the general elections. The following phase is the passcode and the public key sharing phase. This is followed by the voting phase and, finally, the counting phase. The next phase used in this system is the voting phase. This is a crucial phase of the entire process. The confidentiality of every casted vote has to be preserved. This system uses homomorphic encryption whereby upon encryption, and after traversing the network, the ciphertext need not be decrypted in order for the casted vote to be assigned to the appropriate candidate. Figure 21.17 shows a detailed data flow and the result decryption process where it describes how the system achieves this. Figure 21.18 shows a sample interface of what the user would see as they attempt to login into the system. After doing so, the voter would be given the option to cast their votes for their preferred candidate. The eventual results will be displayed into the voters’ account upon reaching the polling day’s end. All the casted votes have been accounted for and verified before it is made available. Drawbacks of this system include performance deficiencies Voter Login Select Candidate to Vote PT_1−0 PT_2−1 .. PT_m−0 Public Key

Encryption Ciphertext Say CT_1−2980 CT_2−1345 .. CT_m−2323 Web Server DB Voter ID 1 2 .. .. n

Candidate 1 2980 CT .. .. CT

Candidate 2 1345 CT .. .. CT

... -CT .. .. CT

Total_A=A[1]*A[2]*----------*A[n]; A=Decrypt(Total_A)/Total number of votes obtain by A Total_B=B[1]*B[2]*----------*B[n], B=Decrypt(Total_B)/Total number of votes obtain by C Total_C=C[1]*C[2]*----------*C[n], C=Decrypt(Total_C)/Total number of votes obtain by C

Candidate m 2323 CT .. .. CT

Figure 21.17  Voting process and winner determining algorithm [6].

Figure 21.18  Mobile website interface samples [6].

416  The Smart Cyber Ecosystem for Sustainable Development and the inability to run ad hoc queries. Homomorphic systems also require application modification or, in some instances, specialized client-server applications.

21.2.1.5 An Online Voting System Based on a Secured Blockchain This research paper’s proposed online voting system intends to use blockchain to develop this system [7]. The first element in the blockchain would be the candidates standing for a position in the elections. This block contains the candidate information, and it would serve as a foundation block. Every vote cast to this particular candidate will be placed on it. This system facilitates a protest vote in which voters may cast a blank vote if they are not satisfied with certain proceedings and wish to express their displeasure. The blockchain is updated each time a voter casts a vote for a particular candidate. Figure 21.19 shows a clear description of this operation and method used to obtain the final results regarding the vote count for every particular candidate contesting in a particular constituency. To guarantee the system would be more secured, the blockchain will contain previous block information, which is also the previous voter’s information. Tampering with these blocks is not feasible because they are all interconnected. It would be easy to detect if any blocks have been tampered with. The blockchain acts centralized, and there should not be a point within the blockchain that fails, which is where the voting occurs. Every incremented vote is just another node addition into the blockchain. Figure 21.20 shows the proposed system as a whole and how voters would interact with the system and methods used by the

Candidate 1

vote 1

vote 2

vote 3

vote n

Candidate 2

vote 1

vote 2

vote 3

vote n

Candidate n

vote 1

vote 2

vote 3

vote n

Figure 21.19  Blockchain structure used by the proposed system [7].

(1) voter

Pc/Device

Interface

(2) Database of registered voters

(3) Encryption

(4)

Figure 21.20  Proposed system framework using blockchain [7].

Blockchain

Secured Online Voting System Using Blockchain  417 systems to store and generate results. This system’s advantages include maintaining voter anonymity and verifying authentication by evaluating the state of the blockchain. This also contributes to the system’s ability to verify each casted vote. This proposed system’s drawbacks include the system is not secured from user devices that have already been exploited by attackers. Furthermore, every voter can only cast his or her vote once, and a mistake is irreversible.

21.2.1.6 Online Voting System Using Fingerprint Biometric and Crypto-Watermarking Approach This research paper proposes an online voting system that utilizes both biometric and crypto-watermarking approaches [8]. This system is divided into three layers, the pre-electoral, electoral, and post-electoral phases. The first phase in this proposed system is the authentication phase. This system uses fingerprint biometric to do so. A fingerprint authentication module is required for voter authentication. A voter registers his or her account by linking the voter fingerprint to a particular registered account. After that, only that fingerprint biometric will be allowed to login into that account. The account registration page is shown in Figure 21.21. The second phase of this proposed system is the voting phase. In this phase, voters would cast their votes for their preferred candidates. This is shown in Figure 21.22. Upon casting their votes, the system uses a qualitative evaluation approach. The system uses anti-watermarking detection tools to evaluate any viable differences between the original image and the watermarked object. This is done using the hash value of the image. If the image has been tampered with, the hash value will differ. Sophisticated image processing software is used to decipher and confirm the hidden message, which is the voter’s casted vote. There will not be any apparent differences in the naked eye because the system utilizes invisible ink embedded in the picture. This means the only way of detecting the difference is by using the image hash value. The final phase of this proposed system is the voting counting phase or the post-electoral phase. After verifying the hash value of an image, all legitimate votes cast would be scrambled before going through the watermarking process. The image hashing process would ensure that all verified votes belong to a legitimate account or a particular voter, as shown

Figure 21.21  Voter registration page [8].

418  The Smart Cyber Ecosystem for Sustainable Development

Figure 21.22  Voting page [8].

Figure 21.23  Hash value of the image with the secret ink [8].

in Figure 21.23. This secures the system from phantom voters or duplicate votes. This system also ensures the confidentiality of the voter. The verified votes are scrambled, making it impossible to trace back the casted vote source. The limitation of this proposed system includes overlooking advancements made in image processing. Attackers these days can utilize various semantics that is making it very much indistinguishable. Another limitation of this system is the threat of collusion attacks. This happens if an image is watermarked multiple times. This inconsistency may be an issue for such online electoral systems.

21.2.1.7 Online Voting System Using Iris Recognition This particular research paper emphasizes the use of iris recognition to enhance the voting system [9]. This system uses the iris image of each voter as the input data into the online voting system. Eligible voters are registered using this iris image. During the login procedure, voter ID is one of the inputs if the password follows the system. The voter’s iris image attempting a login is compared to the iris image of that particular voter’s account stored in the database. If the iris image matches, login is granted to the user of that particular account and would be redirected to the voting panel as shown in Figure 21.24. This

Secured Online Voting System Using Blockchain  419 Voter ID Account

Retina Image & Secret key

Pattern Matching

Verification

Voting Panel

Figure 21.24  Login procedures of the proposed system [9].

particular system emphasizes more on authentication of the voter compared to the confidentiality of the casted vote. Iris recognition is a form of biometric authentication. To register an iris image to a voter’s account, iris extraction procedures have to take effect. Firstly, an image of the eye is taken, followed by the segmentation of the iris. To improve the accuracy of the system, the iris normalization process would help to achieve that. These procedures are taken into account to improve the accuracy of scanning the iris image. The steps are shown in Figure 21.25. At the back end of his proposed system, sophisticated image processing software is used to authenticate the voter. It was discussed previously how the iris image is first registered and linked to a particular voter’s account and also how the system verifies the attempted login to be a legitimate attempt by an authenticated voter who owns the account. The Eye Image

Iris segmentation Iris Normalization Apply wavelet packet transmission

Find energy from all bands

2

Figure 21.25  Iris image registration process [9].

420  The Smart Cyber Ecosystem for Sustainable Development

Figure 21.26  Iris image authentication process [9].

registered iris image linking to that particular account is compared to the voter’s iris image who is attempting to login into that particular account. A match in these two iris images would permit the user to access that particular account. A layout of how this proposed online voting system authenticates a voter is shown in Figure 21.26. Among the strengths of this system includes improved identification accuracy rates with a low risk of a mismatch. Iris identification is a more reliable biometric authentication as it is not affected by environmental factors. The drawback of this proposed system is that systems do not emphasize voter confidentiality. No encrypting algorithms are involved in maintaining the secrecy of the casted vote. Attackers would find it difficult to gain access to a legitimate voter’s account; however, if they succeed in doing so, manipulating the activities carried out by that particular account would be immensely easy to do.

21.2.1.8 Online Voting System Based on NID and SIM These research papers emphasize the use of the NID number and Subscriber Identification Module (SIM) card [10]. The proposed system consists of four phases: the online registration phase, the voting phase, the vote-counting phase, and the results outcome phase. During the first phase of this system, voters would need to register their account with all the necessary information. This system requires every eligible voter to have a SIM card. To register for a sim card, this information would be necessary as well. As soon as an account is created, the account is linked together with the SIM card code that contains the same information as the information registered onto the online voting system. The server encrypts this information, generates a unique NID number for the voter, and is sent to the voter’s email address. Now, the SIM card number and the NID number are combined to form a pin number. All this data is then encrypted. The online registration process is shown in Figure 21.27. The second phase is the voting phase. Every registered and verified account will be granted access to cast their vote for the preferred candidate in this phase. Upon a successful registration phase, the user would need to login in order to cast their vote. The server detects attempted login and sends a notification to the voter’s SIM card, containing the PIN needed to be inserted into the voter system to cast their vote. When inserted into the system, the serves compares the PIN and, if it matches, would grant access to the voter to login into the system. This is shown clearly in Figure 21.28.

Secured Online Voting System Using Blockchain  421

Stores info in Encrypted form

Register Server

Voter

Database

NIC & PIN will be sent to voter’s e-mail Online Registration

Figure 21.27  Voter NID and SIM registration phase [10].

login to the system Server Voter

Voter Verification

Detects SIM & sends notification

Database Login Phase

Figure 21.28  SIM login verification phase [10].

The third phase is the voting phase. The candidate would be redirected to the constituency they belong to. This page would consist of all the candidates contesting in that constituency. The voter would have the option to cast their vote for their preferred candidate. When the vote is cast, the message is encrypted. At the server-side, the message would be decrypted using the symmetric user key. The vote count for the selected candidate would be incremented at the database. This process is shown in Figure 21.29. The final phase of this system is the results announcing phase. This phase does not take effect until the polling day draws to a close. Restriction time is set in the server. During active polling hours, the votes cast are encrypted and decrypted when the polling hours are over. After that, the accumulated votes for every candidate are accounted for, and the outcome of the results will be made available to the registered voters. Figure 21.30 shows the procedures in this phase.

display nominee list Server

Voter

Voter votes & receive notification

perform changes

Database Online Voting

Figure 21.29  Preferred candidate vote count incrementing [10].

422  The Smart Cyber Ecosystem for Sustainable Development display result Server voter’s logout Voter

fetch result of nominee Database

Display Result

Figure 21.30  Results outcome at the end of polling [10].

Figure 21.31  Error messages during an unsuccessful login [10].

Figure 21.31 shows the error message that would be displayed during an unsuccessful login. This is to make sure no unauthorized user would gain access to the system. Among the advantages of using this proposed system includes the integrity of the votes. As long as the polling is active, every casted vote would be encrypted. The system accuracy is also among the strengths as the ballots can be precisely calculated as the duplicate voter is not possible in this system. Finally, voter authentication is very stern in this system, making impersonation very difficult. This system’s drawbacks include the fact that every voter would need to have their own SIM cards. Besides that, a casted vote cannot be withdrawn as it is encrypted immediately. The system is not user error-tolerant.

21.2.1.9 Online Voting System Using Image Steganography and Visual Cryptography The research paper focuses on developing an online voting system using Steganography and Visual Cryptography [11]. This research paper uses a Java EE Application Server to develop

Secured Online Voting System Using Blockchain  423 Server PC

Client Machine

Java EE Application Server (Glassfish) Web Controller

Client (Web-browser)

Request

Controller (Service)

EJB Controller Model Session Beans (EJB)

View (ISP pages)

Response

Database Server (MySQL Driver)

Database

Entity Classes (IPA)

Figure 21.32  Image steganography and visual cryptography system framework [11].

an online voting system. Figure 21.32 shows the framework of the proposed system. This proposed system also consists of four phases: the registration phase, authentication phase, voting phase, and tallying phase. The first phase of this proposed system is the voter registration phase. In this phase, the voter must provide all relevant information to themselves to create an account. This includes a user password. This information added to the system has to be protected. To ensure this, the system adds an extra layer of security as the voter’s password would be cryptographically secured using a dedicated hashed-based scheme. The password salt value is then stored in the system database. Figure 21.33 shows a clear description of this process. The next phase of this system is the voter authentication phase. In this phase, every registered voter account will be authenticated. Voters are required to display the account

: BallotManager

em: javax.persistence.EntityManager

1: addEligibleVoter() 1.1:

customer:Voter

1.2: setName() 1.3: setIcPassport() 1.4: setEmail() 1.5: setOfficer() 1.6: persist(customer) 1.7:

Figure 21.33  User registration phase and password hashing [11].

424  The Smart Cyber Ecosystem for Sustainable Development ID as well as the password. This information would be encrypted for security purposes. The encrypted ciphertext is compared to the ciphertext stored in the database to authenticate a voter or an administrative account. The system does not store the voter or administrator’s passwords. The hashed and salt values of a particular voter of the administrator are stored in the database. This is shown in Figure 21.34. Once the voter is successfully authenticated, they would be redirected to the voting home screen. An example of a homepage available after a successful login by the administrator is shown in Figure 21.35. The next phase of this system is the voting phase. Authenticated voters are allowed to cast their votes by selecting their preferred candidate for the seat that is being contested. Users are allowed to review, reset, and re-select even after the vote is cast. The voter’s : PasswordHash

goodHash: String

Integer : Integer

1: validatePassword() 1.1: split(”:”) 1.2: parseInt(params[ITRERATION_INDEX]) 1.3: fromHox() 1.4: fromHox() 1.5: pbkdf2() 1.6: slowEquals() 1.7:

Figure 21.34  Voter and administrator authentication using hash and salt value [11].

Figure 21.35  Administrator homepage [11].

Secured Online Voting System Using Blockchain  425 final decision and the preferred candidate are selected and image steganography algorithm is used. The preferred candidate is encrypted using stego image. As discussed earlier, this stego image serves as an envelope, which hides the secret message that is intended to be sent. The ballot would be sent to the tally server, where it would be encrypted using a visual cryptography algorithm where shares of the secret message would be generated. This is done as an additional form of security. After reaching the end of the polling day, at the server end, the ballot is decrypted. Finally, we would reach the tallying phase. In this phase, all the decrypted ballots are accounted for, and a winner is declared. Among the strengths of this system is that the system is user error-tolerant. Until the ballot is generated for encryption, users can make changes to their preferred candidate. The system preserves voter’s ballot integrity. This is because of the dual encryption algorithm used in this system. Among the drawbacks of this system are the similar vulnerabilities associated with visual cryptography. Attackers would find it very difficult to decrypt the casted vote; however, it would not be too difficult to manipulate shares that make up a ballot causing inconsistent polling results.

21.2.1.10 Online Voting System Using Secret Sharing–Based Authentication This research paper uses secret sharing–based authentication to build a secured online voting system [12]. This proposed system consists of four phases. They are the registration phase, authentication phase, vote casting phase, tallying phase, and audit trail phase. In the first phase, eligible voters would need to register themselves into the system. The system would generate the password. Once voter registration is successful, the password is delivered to the voter using Shamir’s (2, 2) threshold secret sharing scheme. The password is sent in shares, with each share encrypted using this algorithm. Figure 21.36 shows the operational procedures of this proposed system. The next phase of this proposed system is the authentication phase. In this phase, the shared secret is efficiently used for authentication purposes. The voter will have to produce their share at this phase if the administrative system produces its share. Both these shares are crucial if the system intends to reconstruct the initial secret key. A reconstruction model is generated to merge the shares for authentication purposes. Once the voter has been authenticated, they would proceed to the next phase, the voting phase. Upon this point, the procedures become standard as the voter can cast their vote for their preferred candidate. At the end of the polling day, all the casted vote is accounted for, and the winner is announced. Figures 21.37 and 21.38 show the algorithm to form shares and the algorithm to reconstruct them at the receiver’s end. Among the strength of this system is the use of the secret sharing scheme. The shares produce in this system is difficult to be decrypted by an attacker making impersonation in this proposed system very difficult. However, the weakness remains that an attacker may manipulate or add shares that would corrupt the key at the receiver’s end.

426  The Smart Cyber Ecosystem for Sustainable Development Voter and candidate registration Trusted Center

Authentication

Authority Vote casting Tallying Audit Trial Figure 21.36  Voting system based on secret sharing authentication framework [12].

Algorithm to form shares Input: d is secret in the form of an integer, n is number of participants, and k is threshold Output: shares for the n participants to keep. Step 1. Choose randomly a prime number p that is larger than d. Step 2. Select (k-1) integer values c1, c2..... ck-1 within the range of 0 to p-1 Step 3. Select n distint real values x1,x2.....xn Step 4. Use the following (k-1)-degree polynomial to compute n function values .F(xi) called partial shares for i=1,2...n, F(xi)=(d+c1xi+c2xi2....+ck-1xik-1) mod p Step 5. Deliver the twp-tuple (xi.F(xi))as a share to the ith participant where i=1,2...n.

Figure 21.37  Algorithm to form shares [12].

Algorithm for reconstruction Input: k shares collected from the n participants and the prime number p. Output: secret d hidden in the shares. Step1. Use k shares (x1.F(x1)).(x2.F(2x))......(xk.F(xk)) To from F(xi)=(d+c1xi+c2xi2....+ck-1xik-1) mod p where i=1.2.....k. Step 2. Solve the k equations above by Lagrenge’s interpolation to obtain as follows d=(-1)k-1[F(x1)*(x2x3.....xk)/((x1-x2)(x1-x3)....(x1xk))+F(x2)*(x1x3)......xk-1)/((x2-x1)(x2-x3)......(x2-x3))..... +F(xk)*(x1x2.....xk-1)/((xk-x1)(xk-x2)....(xk-xk-1))] mod p Here d is the original secret password . Voter can cast vote only if the recovered Password match with the original secret.

Figure 21.38  Algorithm to reconstruct shares [12].

Secured Online Voting System Using Blockchain  427

21.2.2 Comparing the Existing Online Voting System Table 21.1 presents a comparison of the existing online voting systems in the literature. Table 21.1  Comparing the existing online voting system. Description of proposed system

Strengths of the proposed system

Weaknesses of the proposed system

1. Online Voting System Based on Cryptographic and Stego-Cryptographic Model [3]. The authors of this journal used steganographic and cryptographic algorithms to develop a secured online voting system. Steganographic systems would consist of a secret message, cover data, and the stego message.

It has the ability to encode secret messages using various forms of data cover and also the ability for the system to camouflage hidden data as it would appear the same as the input data, thus making it difficult to be identified.

The proposed system requires large data to be transmitted in order for this algorithm to work. This is because the system uses unused bits of a larger data to hide or envelope the secret message that needs to be sent. Therefore, the larger the secret message, even larger data is required to aid this process.

2. Online Voting System Based on Visual Cryptography [4]. This system achieves this by generating shares of visual cryptography using the basic cryptographic model.

This proposed system is well secured from attackers who wish to decrypt the information traveling in the network.

This system has potential weaknesses as attackers may significantly affect the content of shares in the system.

3. Online Voting System Using Biometric Security and Steganography [6]. Biometric identity is used in this system for added security purposes. Steganography is used in this system as it serves as a media cover.

The use of biometric measures in ensuring that the online voting system is fully secured does give a significant advantage in terms of the security strengths of this system.

The drawback of this system however, would be the expansion of the secret key. Due to large number of users, a long secret key is required. This may not be good for voters who are not able to memorize the secret key.

4. Cloud-Based Secured Online Voting System using Homomorphic Encryption [6]. This system utilizes Paillier’s encryption technique. Homomorphic encryption can be very beneficial as operations that use the stored data to identify the winner of the polling without having the need to decrypt the data.

This system uses homomorphic encryption whereby upon encryption, and after traversing the network, the ciphertext need not be decrypted in order for the casted vote to be assigned to the appropriate candidate.

Drawbacks of this system include performance deficiencies and inability to run ad hoc queries. Homomorphic systems also require application modification or in some instances specialized clientserver applications.

(Continued)

428  The Smart Cyber Ecosystem for Sustainable Development Table 21.1  Comparing the existing online voting system. (Continued) Description of proposed system

Strengths of the proposed system

Weaknesses of the proposed system

5. An Online Voting System Based on a Secured Blockchain [7]. The proposed online voting system in this research paper intends on using blockchain to develop this system.

The advantages of this system include maintaining voter anonymity and ability to verify authentication by evaluating state of the blockchain. This also contributes to the system ability to verify each casted vote.

Drawbacks of this proposed system include the system is not secured from user devices that has already been exploited by attackers. Furthermore, every voter can only cast his or her vote once and a mistake is irreversible.

6. Online Voting System Using Fingerprint Biometric and CryptoWatermarking Approach [8]. This research paper proposes an online voting system that utilizes both biometric as well cryptowatermarking approaches.

This secures the system from phantom voters or duplicate votes. This system also ensures confidentiality of the voter. The verified votes are scrambled thus making it impossible to trace back the source of a casted vote.

Limitation of this proposed system includes overlooking advancements made in image processing. Attackers these days are able to utilize various semantics that are making it very much indistinguishable. Another limitation of this system is the threat of collusion attacks. This happens if an image is watermarked multiple times. This inconsistency may be an issue for such online electoral systems.

7. Online Voting System Using Iris Recognition [9]. This particular research paper emphasizes on the use of iris recognition to enhance security measures of the voting system. This system uses the iris image of each voter as the input data into the online voting system.

Among the strengths of this system is the improved identification accuracy rate with low risk of mismatch. Iris identification is a more reliable biometric authentication as it is not affected by environmental factors.

Drawback of this proposed system is that systems do not emphasize on voter confidentiality. No encrypting algorithms are involved in maintaining the secrecy of the casted vote. Attackers would find it difficult to gain access to a legitimate voter’s account, however, if they succeeded in doing so, then manipulating the activities carried out by that particular account would be immensely easy to do. (Continued)

Secured Online Voting System Using Blockchain  429 Table 21.1  Comparing the existing online voting system. (Continued) Description of proposed system

Strengths of the proposed system

Weaknesses of the proposed system

8. Online Voting System Based on NID and SIM [10]. This research papers emphasizes the use of National Identification (NID) number and Subscriber Identification Module (SIM) card.

Among the advantages of using this proposed system is the integrity of the votes. As long as the polling is active, every casted vote would be encrypted. The system accuracy is also among the strengths as the ballots can be precisely calculated as duplicate voter is not possible in this system. Finally, voter authentication is very stern in this system making impersonation very difficult.

Drawbacks of this system include the fact that every voter would need to have their own SIM cards. Besides that, a casted vote cannot be withdrawn as it is encrypted immediately. The system is not user error-tolerant.

9. Online Voting System Using Image Steganography and Visual Cryptography [11]. The research paper focuses on developing an online voting system using steganography and visual cryptography. This research paper uses a Java EE application server to develop an online voting system.

Among the strengths of this system is that the system is user error-tolerant. Until the ballot is generated for encryption, user can make changes into their preferred candidate. The system preserves voter’s ballot integrity. This is because of the dual encryption algorithm used in this system.

Among the drawbacks of this system are the similar vulnerabilities associated with visual cryptography. Attackers would find it very difficult to decrypt the casted vote; however, it would not be too difficult in manipulating shares that make up a ballot causing inconsistent polling results.

10. Online Voting System Using Secret Sharing– Based Authentication [12]. This research paper uses secret sharing–based authentication in order to build a secured online voting system.

Among the strengths of this system is the use of the secret sharing scheme. The shares produce in this system is difficult to be decrypted by an attacker making impersonation in this proposed system very difficult.

The weakness however, remains that an attacker may manipulate or add shares that would corrupt the key at the receiver’s end.

430  The Smart Cyber Ecosystem for Sustainable Development

References 1. Sofian, N., Wicaksana, A., Hansun, S., LSB steganography and AES encryption for multiple PDF documents. 2019 5th International Conference on New Media Studies (CONMEDIA), IEEE, pp. 100–105, 2019. 2. Budianto, C.D., Wicaksana, A., Hansun, S., Elliptic curve cryptography and lsb steganography for securing identity data. International Conference on Applied Computing and Information Technology, Springer, Cham, pp. 111–127, 2019, May. 3. Olaniyi, O.M., Tayo, A.O., Olusayo, O.E., Olusola, O.O., A survey of cryptographic and stegano-­ cryptographic models for secure electronic voting system. Covenant J. Inf. Commun. Technol., 1, 2, 54–78, 2013. 4. Raj, G., Jithamol, P.M., Narayanan, N., Online Voting: Using Visual Cryptography. IJSRSET, 2, 3, 2016. 5. Katiyar, S., Meka, K.R., Barbhuiya, F.A., Nandi, S. Online voting system powered by biometric security using steganography. 2011 Second International Conference on Emerging Applications of Information Technology, IEEE, pp. 288–291, 2011, February. 6. Ranjan, M., Mondal, A.H., Saikia, M., A Cloud Based Secure Voting System using Homomorphic Encryption for Android Platform. Int. J. Electr. Comput. Eng. (IJECE), 6, 6, 2994–3000, 2016. 7. Ayed, A.B., A conceptual secure blockchain-based electronic voting system. Int. J. Network Secur. Its Appl., 9, 3, 01–09, 2017. 8. Olaniyi, O.M., Folorunso, T.A., Aliyu, A., Olugbenga, J., Design of secure electronic voting system using fingerprint biometrics and crypto-watermarking approach. Int. J. Inf. Eng. Electr. Bus., 8, 5, 9, 2016. 9. Nithya, Ms. J, Abinaya, G., Sankareswari, B., Saravana Lakshmi, M., Iris recognition based voting system. International Conference on Science, Technology, Engineering & Management [ICON-STEM’15], 2015. 10. Ghate, B., Talewar, S., Taware, S., Katti, J.V., E-Voting System based on Mobile using NIC and SIM. Int. J. Comput. Appl., 165, 8, 9–13, 2017. 11. Rura, L., Issac, B., Haldar, M.K., Online voting system based on image steganography and visual cryptography. J. Comput. Inf. Technol., 25, 1, 47–61, 2017. 12. Walake, M.A. and Chavan, M.P., Efficient Voting system with (2 2) Secret Sharing Based Authentication. (IJCSIT) Int. J. Comput. Sci. Inf. Technol., 6, 1, 410–412, 2015.

22 Artificial Intelligence and Cybersecurity: Current Trends and Future Prospects Abhinav Juneja1, Sapna Juneja2*, Vikram Bali3, Vishal Jain4 and Hemant Upadhyay5 KIET Group of Institutions, Ghaziabad, India 2 IMS Engineering College, Ghaziabad, India 3 JSS Academy of Technical Education, Noida, India 4 Sharda University, Greater Noida, India 5 BMIET, Sonepat, India 1

Abstract

Cybersecurity is a wide term and it is an area which relates with various organizations and governments, each at different level, usually from solitary to country wise. So, Artificial Intelligence (AI) along with Machine Learning (ML) techniques is being used over the spectrum of security. Most of these technologies have their own future prospects and are providing the security to the community by reducing the frauds in digital transactions, etc. Cybersecurity and AI can provide fruitful results if run in parallel, and the current approaches of ML appears to be more perfect in fulfilling the breaches of earlier rule-based security structures. AI helpfully accumulate, sort, and examine enormous amount of data and permits organizations to extract additional utility of the data. These attributes have been used in cybersecurity sector. The purpose of this research is to highlight ongoing trends and applications to achieve cybersecurity using AI at organizational level. Keywords:  Cybersecurity, deep learning, KNN, machine learning, malware, neural networks, NLP, shallow learning

22.1 Introduction Cybersecurity can broadly be represented as formulation of shielding approach which is used to protect computational assets, devices, data, and networks from illegal access, modification, and shattering [1]. Because of the frequent upgradations of the communication methods, more and more network security threats are evolving and changing swiftly. Hackers are trying to adopt novel and advanced methods that can speed up the range of the attack. So, there is a need of developing more workable, versatile, and vigorous system of cyber resistance that works in real time environment in successful identification of cyber attacks [2]. The usage of Artificial Intelligence in cybersecurity for prevention and detection *Corresponding author: [email protected] Pardeep Kumar, Vishal Jain and Vasaki Ponnusamy (eds.) The Smart Cyber Ecosystem for Sustainable Development, (431–442) © 2021 Scrivener Publishing LLC

431

432  The Smart Cyber Ecosystem for Sustainable Development of hack has been increased. Apart from the usage of AI in Healthcare, Intelligent Games, Computer Vision, NLP, and education, AI is influencing network security for identification of cyber threats [3]. Researchers have tried to present various reviews in the field of cybersecurity and AI. Some has used Machine Learning methods while others have used Deep Learning for cybersecurity [4].

22.2 Literature Review Apruzzese et al. [5] defined various ML and DL methods to protect from cyber attacks. The approaches which they were defined were capable of identifying malware attacks, diagnosis of spam mail, and network intrusion diagnosis. Li [6] highlighted the combination of AI along with ML. Researcher defined some approaches used to deal with cyber attacks. The author mentioned the likelihood of phishing with the AI model. Xin et al. [4] presented the differences of ML and DL methods that can be employed for cybersecurity. They focused on intrusion detection technique only in order to protect the network. Xu et al. [7] used ML technique to develop a structure that has been used for hardware-supported malware diagnosis that works on usage patterns of virtual memory when becomes online. The algorithms which they used were SVM, RFC, and Logistic Regression technique. They found an accuracy of around 99% in detecting the cyber attacks. Hashemi et al. [8] used various ML classifiers to categorize the malwares. The classifiers which they implemented are KNN, SVM, and OPCODE. They proved with their results that, this model is capable of detecting the malware at a higher rate. Mclaughlin et al. [9] highlighted a CNN model for the diagnosis of malware. The program had been split into a series of opcode which was used to define the type of malware. Ye et al. [10] developed a DL-based model for detection of malwares. The structure defined was comprised of Multilayer Boltzmann Machine along with auto-encoder to detect the intruders. Wang et al. [11] in this work, we propose a hybrid model based on deep autoencoder (DAE used a combinational model of CNN and Deep auto-encoder to improvise the performance of detection of malware attacks but this system was applicable only on Android-based applications. Al-Yaseen et al. [12] used SVM and DL along with K-means algorithm for the intrusion detection system. Their model was capable of achieving an accuracy of 95.75% with 1.87% of failure rate. Kabir et al. [13] defined a technique for detecting cyber attacks using sampling method using least square support vector machine. KDD cup dataset was used and the system was much efficient for intrusion detection.

22.3 Different Variants of Cybersecurity in Action Cybersecurity [14] however, is time-consuming and requires in-depth security domain knowledge. Moreover, large amounts of labeled legitimate and attack request data are needed by supervised learning algorithms to classify normal and abnormal behaviors, which is often expensive and impractical to obtain for production web applications. This paper provides three contributions to the study of autonomic intrusion detection systems. First, we evaluate the feasibility of an unsupervised/semi-supervised approach for web attack detection based on the Robust Software Modeling Tool (RSMT) refers to a group of

Artificial Intelligence and Cybersecurity  433 Cybersecurity

Network Security

Cloud Security

Application Security

IoT Security

Figure 22.1  Variants of cybersecurity.

protocols which any organization or any person adopts to confirm the confidentiality of its data. It basically keeps any organization protected from various types of cyber attacks. The basic goal of cybersecurity has been split into two parts as failure recovery and business progression [15]. Failure recovery refers to maintain the unification of the information even after any attack and business progression means retaining the data and continuing the business even in difficult conditions [16]. The variants of cybersecurity [17] that are in action nowadays are represented by Figure 22.1. Network Security [18] secures any organization from illegal usage or intrusion. If the network is completely secured, then it can locate and remove threats inside the organization as well. But the problem associated with implementation of network security is that it requires more number of resources and thus reduces the overall productivity of the organization. Cloud Security includes the methods, approaches, and techniques acquired to secure the cloud-based system and its associated data. Cloud security provides special measures that are used to secure the privacy and data of the users. Application Security is to secure all the applications of the organization from the intruders. Usually application security is provided in the development phase of the application, but some other tools have also been used to enhance the security once the application is ready to use. IoT Security is associated with online communication. It is usually deployed for devices and networks which are connected with the IoT [19]. In IoT, when devices connect with each other for data sharing, it becomes open for attackers to attack, if not secured properly [20].

22.4 Importance of Cybersecurity in Action The areas where cybersecurity is much important are government organizations, armed forces, business organizations, health organizations, and financial departments in order to accumulate, use, and pile up exceptional data on the systems [21]. If the data has some sensitive information that can be used by some unauthorized agencies that can cause very adverse effects [22]. Organizations broadcast important and sensitive information to other systems and networks for their business purpose and then comes the role of cybersecurity which is to protect this important information from unauthorized access by imposing various methods and techniques [23]. As in today’s scenario, cyber attacks are increasing day by day; organizations who are dealing with the important data related to the nationalized defense, records of finance departments, etc., are taking necessary steps and precautions to safeguard their data from intruders and to control terrorism [24].

434  The Smart Cyber Ecosystem for Sustainable Development

22.5 Methods for Establishing a Strategy for Cybersecurity Every organization must maintain a strategy for ensuring the cybersecurity which should be followed by the whole organization. There are some techniques that can be used to maintain the strategy of cybersecurity in the organization as follows. 1. Carry out a record of all the computable belongings of the organization: First of all, maintain a list to explore that which information and applications are being used by the organization and what would be the consequence if that data or application will be hacked by some intruder [25]. So, first of all, a list of such data or applications is needed to be created in order to protect it. 2. Recognize any compliance needs, if any. It is needed to be verified whether the organization is following some rules or laws that are affecting cybersecurity of the organization [26] which bridge the physical and virtual worlds. We propose a new taxonomy of attacks, which classifies them into four broad categories. The most interesting category which we call functionality extension attacks. So, if there are such rules, then these rules also must be added to the above mentioned record. 3. Recognize and categorize the threats involved: Make a list about the possible available threats that can affect the organization and which threats are more hazardous [27] smart environment, e-healthcare, etc. Due to the presence of unreliable internet and new routing protocols for low-power devices, IoT requires innovative security solutions. In this paper, we present three new Intrusion Detection Systems (IDSs). 4. Generate the solutions in order to deal with the threats by identifying their severity level starting from the most severe threat for the organization [23]. 5. Classify the available cybersecurity program of the organization. Check whether there is sufficient manpower to handle with security issues. Figure out the preexisting cybersecurity methods [28]. 6. Create a specific team to handle only cybersecurity related issues. Train the staff in order to deal with cybersecurity related problems, hire new staff if required, and try to minimize the cyber crime–related issues within the organization [29]. 7. Create some timelines and follow them in routine to check whether the cyber system of the organization is proper or not [30].

22.6 The Influence of Artificial Intelligence in the Domain of Cybersecurity The main purpose of AI is to apply the intelligence to generate smart machines or devices which are capable of self-learning and decision making [6]. The main idea was to generate the machines that behave like human beings. Another major use of AI is to solve complex problems that can be resolved either in real time environment or based upon decisionmaking [31] we have seen that Artificial Intelligence (AI). As AI is emerging in many fields

Artificial Intelligence and Cybersecurity  435 like Computer vision, Speech Recognition, Natural Language Processing, and DL, but its growth has created many issues related to cybersecurity [32]. These issues are as follows: 1. There is a boom of enormous data which is difficult to handle manually [5]. 2. Cyber threats are increasing rapidly and become more adaptive even to the difficult environments [2]. 3. It becomes very difficult to differentiate between ordinary data and a file containing contagious data [33]. 4. The process of preventing these threats is very time- and money-consuming [34]. So by looking at the above issues, it is expected that emerging AI technologies should be able to deal with these cybersecurity issues. So, researchers have put their efforts in order to generate various ML algorithms for dealing with cybersecurity [33]. These algorithms [5] shown by Figure 22.2 has been categorized as follows. The first level of ML algorithm is divided into two parts: Shallow Learning and Deep Learning. Shallow Learning [10] is a type of ML in which first feature, which are problemdependent, are extracted first and then model has been trained according to that features while in DL methodology [6], both the tasks of feature extraction and training of model is done at the same time. Both the techniques are again divided into two parts as Supervised and Unsupervised Learning. Supervised Learning [10] performs the conversion of the input into desired pre-identified output while unsupervised learning, is not aware about the output data and is not aiming toward generation of desired data, rather it tries to identify patterns in the output data. Supervised Shallow Learning is partitioned into various algorithms like Naïve based, Logistic Regression, SVM, Random Forest, KNN, and Neural Networks.

Machine Learning

Shallow Learning

Supervised Shallow Learning

Deep Learning

Unsupervised Shallow Learning

Supervised Deep Learning

Naïve Based

Clustering

Fully Connected DNN

Stack Auto Encoders

Logistic Regression

Association

Convolutional DNN

Deep Belief Networks

Support Vector Machines

Recurrent DNN

Random Forest K Nearesy Neighbor Neural Networks

Figure 22.2  Categories of machine learning methods for cybersecurity.

Unsupervised Deep Learning

436  The Smart Cyber Ecosystem for Sustainable Development Naïve-based algorithm [35] is based upon the Bayes Theorem of conditional probability. This theorem presumes that attributes or inputs are independent of each other. This model requires little data and provides fast results with better accuracy. Further, this model is easily scalable. In Logical Regression Model [36], the output function can consider only distinct value for a given set of input functions. The performance of logical regression model is dependent upon volume of the training data. Next in the queue are Support Vector Machines. These are the types of classifiers [37] which do not create any type of assumption on the input data set and performs the mapping of samples of data by maximizing the space between patterns of samples. Usually their processing time remains very long. Random Forest Learning Algorithm [18] is used for regression and classification also. The Random Forest Algorithm performs the task of decision-making by generating multiple decision trees of the set of available conditions and then generates a prediction based upon that decision trees and finally choose the most optimizing solution from that decision trees. Next algorithm is KNN [8] which identifies the resemblance between the new data and previously available data set and then add the new data into similar type of data category. KNN is also used both for regression as well for classification. It usually does not make any hypothesis on the dataset. Neural Networks or shallow neural networks is made up of neurons and is arranged using one or two layers. All the neural networks made up of either less number of neurons or layers come under this category. In cybersecurity, shallow neural networks [38] are used for classification. Clustering Algorithm divides the data sets into various groups such that one group contains almost similar type of data but it is totally different from data of some other group. So, it is a grouping of objects depending upon their resemblance. Association is used to recognize hidden patterns between data to make them appropriate for prediction. Association algorithm [37] computes the rate of repetition of same data from a large collection of dataset by aiming to find the associations between data items. Fully connected Deep Neural Networks [39] are a category of Deep Neural Networks in which neurons of consecutive layers are completely connected with each other without making the assumption on the input dataset. It gives a general solution for the classification, whereas Convolutional Neural Networks are the type of Neural Network that comprises of invisible layers. These layers have convolution, pooling, and activation functions. These functions are used for non-linearity. Recurrent neural networks are a type of neural networks in which input is passed through a series of layers to generate the output, with the condition that output generated is dependent upon the preceding inputs. They are most commonly used for Natural Language Processing. Stack auto-encoders [40] are neural networks, made up of various auto-encoders in which input and output neurons are similar in number. In this neural network, the dataset is pre-trained to get the more accurate results. Lastly, the Deep Belief is the type of networks in which the different layers of the network are connected with each other but the units between the layers are not connected with each other. The solution provided by ML [41] in order to secure the data from cyber attacks consists of these major steps as follows. 1. Identify the features. 2. Choose the suitable ML algorithm. 3. Apply the data set to train the model and then identify the best model which gives the best performance by assessing multiple algorithms. 4. Use the trained model to forecast the unknown data.

Artificial Intelligence and Cybersecurity  437

22.7 Where AI Is Actually Required to Deal With Cybersecurity AI’s main role [42] a reliable Next Generation Cybersecurity Architecture (NCSA) is to unload the task of cybersecurity from human engineers and to manage it more precisely and perfectly in the way that human beings cannot do. AI along with ML should be able to adapt the constant changes in malwares and threats and diagnose the problem as soon as it arises. So, here is the need of the cybersecurity [1] which the AI techniques and tools assist to meet. 1. Management of large amount of security data—It is become quite difficult for human beings to handle a huge amount of data that is producing at a very faster rate in today’s world. So, with the help of AI technologies, it becomes very easy to handle this data and hence to send the security alerts on time for providing more and more security [43]. 2. Speeding up the task of detection of anomalies and fast response time—AI can accelerate the identification of any problem by matching the origin of the security data and alerts generated which seems an impossible task for human experts [44]. 3. Easy for AI to pick up the criminal activity among a set of activities—While hackers perform their task of attacking the data, they becomes hidden from the user or human expert but AI can easily locate out some suspicious activity and behavioral pattern to identify whether it is some security threat or not [45]. ML is beneficial for any organization in terms of providing security. Loss of data and other types of attacks that every organization faces is becoming a severe problem nowadays. Organizations are using ML software to enhance their data security, depending upon their need, budget, and size [14] however, is time-consuming and requires in-depth security domain knowledge. Moreover, large amounts of labeled legitimate and attack request data are needed by supervised learning algorithms to classify normal and abnormal behaviors, which is often expensive and impractical to obtain for production web applications. This paper provides three contributions to the study of autonomic intrusion detection systems. First, we evaluate the feasibility of an unsupervised/semi-supervised approach for web attack detection based on the Robust Software Modeling Tool (RSMT). ML techniques which organizations are using can be elaborated as follows. Classification: Classification can be used for filtering out the spam data or emails, to stop phishing, and to detect any malicious thing by using pre-decided parameters [46]. Clustering: Clustering collects the data without any requirement of identification of groups so that the cybersecurity engineer should be able to identify the place of security attack [47] while reflecting the need for enhancing constantly NATO’s (North Atlantic Treaty Organization. Association learning rule: The association learning program observes various activities and actions involved and learn from them. From that learning, it disclose the patterns of risk from various events and then guide the system on how to behave with any specific event and how to deal with the risk associated with that particular event [16].

438  The Smart Cyber Ecosystem for Sustainable Development Prediction Modeling: ML algorithm collects the data and speculates any criminal activity which has been further assisted by the security team to take the proper action before that activity converts into cyber attack [48].

22.8 Challenges for Cybersecurity in Current State of Practice Although the community belonging to ML is not aware, but cybersecurity associated with ML also faces challenges in real time systems. These challenges can be classified as follows. 1. First major challenge of cybersecurity using ML is to diagnose and classify the malware. As the attackers are also aware of latest technology updates, and can impose complex techniques in malware program like encryption and compression to reduce the chances of attack detection, it becomes difficult to diagnose the maleficent program [49]. 2. Another challenge which the community facing is finite expertise in the specific domain that causes trouble in recognizing the source of malware program and hence not aware about the severity of the disaster caused by that particular malware attack [50]. 3. The validity of AI algorithms is an important obstacle. Processing time is wasted using false alarms leading the AI mechanism to pass over the malware attack [51].

22.9 Conclusion AI along with ML is playing an important role in the field of cybersecurity and trying to maximize it. Also these techniques are putting a strong impact on human’s life and trying to improve it. ML-based cybersecurity systems can uncover the threats and security breaches. By observing the correlation between different security attacks for a specific duration, ML algorithm can uncover the risks in security and the possible security attacks for the future. It appears that AI in parallel with ML can provide the complete solution for security threats, attacks, and all kinds of breaches but organizations in the world are not able to handle these threats. So, organizations around the world must try to adopt AI- or ML-based methodologies and tools for securing their businesses. They must also be aware of ML Algorithms, their applicability, their security strengthening mechanism, and methods to train these ML algorithms. Apart from many pros and cons of AI and ML, it is still the brightest approach to deal with cybersecurity related issues. In order to obtain the maximum benefit from AI and ML, the cybersecurity domain must increase their proficiency and grasping of these new technologies.

References 1. Sagar, B.S., Niranjan, S., Kashyap, N., Sachin, D.N., Providing cybersecurity using artificial intelligence - A survey. Proc. 3rd Int. Conf. Comput. Methodol. Commun. ICCMC 2019, pp. 717–720, no. Iccmc, 2019.

Artificial Intelligence and Cybersecurity  439 2. Mittu, R. and Lawless, W.F., Human factors in cybersecurity and the role for AI. AAAI Spring Symp. - Tech. Rep, vol. SS, pp. 39–43, 2015. 3. Mohanty, R., Will AI Change the Game for Cybersecurity in 2018? Will AI Change the Game for Cybersecurity in 2018?, 2018. 4. Xin, Y. et al., Machine Learning and Deep Learning Methods for Cybersecurity. IEEE Access, 6, c, 35365–35381, 2018. 5. Apruzzese, G., Colajanni, M., Ferretti, L., Guido, A., Marchetti, M., On the effectiveness of machine and deep learning for cybersecurity. Int. Conf. Cyber Conflict, CYCON, vol. 2018-May, pp. 371–389, 2018. 6. hua Li, J., Cybersecurity meets artificial intelligence: a survey. Front. Inf. Technol. Electron. Eng., 19, 12, 1462–1474, 2018. 7. Xu, Z., Ray, S., Subramanyan, P., Malik, S., Malware detection using machine learning based analysis of virtual memory access patterns. Proc. 2017 Des. Autom. Test Eur. DATE 2017, pp. 169–174, 2017. 8. Hashemi, H., Azmoodeh, A., Hamzeh, A., Hashemi, S., Graph embedding as a new approach for unknown malware detection. J. Comput. Virol. Hacking Tech., 13, 3, 153–166, 2017. 9. Mclaughlin, N. et al., Deep Android Malware Detection. 301–308. 10. Ye, Y., Chen, L., Hou, S., Hardy, W., Li, X., DeepAM: a heterogeneous deep learning framework for intelligent malware detection. Knowl. Inf. Syst., 54, 2, 265–285, 2018. 11. Wang, W., Zhao, M., Wang, J., Effective android malware detection with a hybrid model based on deep autoencoder and convolutional neural network. J. Ambient Intell. Humaniz. Comput., 10, 8, 3035–3043, 2019. 12. Al-Yaseen, W.L., Othman, Z.A., Nazri, M.Z.A., Multi-level hybrid support vector machine and extreme learning machine based on modified K-means for intrusion detection system. Expert Syst. Appl., 67, 296–303, 2017. 13. Kabir, E., Hu, J., Wang, H., Zhuo, G., A novel statistical technique for intrusion detection systems. Futur. Gener. Comput. Syst., 79, 303–318, 2018. 14. Pan, Y. et al., Detecting web attacks with end-to-end deep learning. J. Internet Serv. Appl., 10, 1, 2019. 15. Demertzis, K. and Iliadis, L., A bio-inspired hybrid artificial intelligence framework for cybersecurity. Comput. Cryptogr. Netw. Secur., September, 161–193, 2015. 16. Engineering, T., Implementation of a Robust and Safe Cybersecurity System By Preventing the Intrusion of Outsiders By Formulation of a Novel and Efficient Cyber Law. IJITMIS, 6, 2, 1–10, 2015. 17. Bendovschi, A., Cyber-Attacks – Trends, Patterns and Security Countermeasures. Procedia Econ. Financ., 28, January, 24–31, 2015. 18. Truong, T.C., Diep, Q.B., Zelinka, I., Artificial intelligence in the cyber domain: Offense and defense. Symmetry (Basel)., 12, 3, 1–24, 2020. 19. Juneja, S., Juneja, A., Anand, R., Role of big data as a tool for improving sustainability for the betterment of quality of life in metro cities. Int. J. Control Autom., 12, 5, 553–557, 2019. 20. Juneja, A., Bajaj, S., Anand, R., Sindhwani, N., Improvising green computing using multicriterion decision making. J. Adv. Res. Dyn. Control Syst., 12, 3 Special Issue, 2020. 21. Čapek, J., Cybersecurity and internet of things. IDIMT 2018 Strateg. Model. Manag. Econ. Soc. 26th Interdiscip. Inf. Manag. Talks, vol. 7, no. 1, pp. 343–349, 2018. 22. Dlodlo, N., Gcaba, O., Smith, A., Internet of things technologies in smart cities. 2016 IST-Africa Conf. IST-Africa 2016, pp. 1–7, 2016. 23. Burhan, M., Rehman, R.A., Khan, B., Kim, B.S., IoT elements, layered architectures and security issues: A comprehensive survey. Sensors (Switzerland), 18, 9, 1–37, 2018. 24. Bugeja, J., Jönsson, D., Jacobsson, A., An Investigation of Vulnerabilities in Smart Connected Cameras. 2018 IEEE Int. Conf. Pervasive Comput. Commun. Work. PerCom Work. 2018, pp. 537–542, 2018.

440  The Smart Cyber Ecosystem for Sustainable Development 25. Bastos, D., Shackleton, M., El-Moussa, F., Internet of things: A survey of technologies and security risks in smart home and city environments. IET Conf. Publ, vol. 2018, no. CP740, 2018. 26. Ronen, E. and Shamir, A., Extended functionality attacks on IoT devices: The case of smart lights. Proc. - 2016 IEEE Eur. Symp. Secur. Privacy, EURO S P 2016, pp. 3–12, 2016. 27. Shukla, P., ML-IDS: A machine learning approach to detect wormhole attacks in Internet of Things. 2017 Intell. Syst. Conf. IntelliSys 2017, vol. 2018-January, no. September, pp. 234–240, 2018. 28. Tiwari, V.K. and Dwivedi, R., Analysis of cyber attack vectors. Proceeding - IEEE Int. Conf. Comput. Commun. Autom. ICCCA 2016, pp. 600–604, no. February, 2017. 29. Archer, E.M., Crossing the rubicon: Understanding cyber terrorism in the european context. Eur. Leg., 19, 5, 606–621, 2014. 30. Park, J. and Tyagi, A., Using Power Clues to Hack IoT Devices: The power side channel provides for instruction-level disassembly. IEEE Consum. Electron. Mag., 6, 3, 92–102, 2017. 31. Chawla, P., Juneja, A., Juneja, S., Anand, R., Artificial intelligent systems in smart medical healthcare: Current trends. Int. J. Adv. Sci. Technol., 29, 10 Special Issue, 1476–1484, 2020. 32. Chen, J., Su, C., Yan, Z., AI-Driven Cybersecurity Analytics and Privacy Protection. Secur. Commun. Networks, 2019, 2019. 33. Tyugu, E., Artificial intelligence in cyber defense. 2011 3rd Int. Conf. Cyber Conflict, ICCC 2011 – Proc., no. January 2011, 2011. 34. Hughes, R., 済無No Title No Title. J. Chem. Inf. Model., 53, 9, 287, 2008. 35. Raschka, S., Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning arXiv: 1811. 12808v1 [cs. LG] 13 Nov 2018, 2018. 36. Peng, C.Y.J., Lee, K.L., Ingersoll, G.M., An introduction to logistic regression analysis and reporting. J. Educ. Res., 96, 1, 3–14, 2002. 37. Evgeniou, T. and Pontil, M., Workshop on Support Vector Machines: Theory and Applications. Mach. Learn. Its Appl. Adv. Lect., January 2001, 249–257, 2001. 38. Min, Y. and Chung, H.W., Shallow Neural Network can Perfectly Classify an Object following Separable Probability Distribution. IEEE Int. Symp. Inf. Theory - Proc, vol. 2019-July, pp. 1812–1816, no. 1, 2019. 39. Dai, D., Tan, W., Zhan, H., Understanding the Feedforward Artificial Neural Network Model From the Perspective of Network Flow, 2017. [Online]. Available: http://arxiv.org/ abs/1704.08068. 40. Liu, G., Bao, H., Han, B., A Stacked Autoencoder-Based Deep Neural Network for Achieving Gearbox Fault Diagnosis. Math. Probl. Eng., 2018, 2018. 41. Thomas, T., Vijayaraghavan, A.P., Emmanuel, S., Machine Learning Approaches in Cybersecurity Analytics. Mach. Learn. Approaches Cyber Secur. Anal., November 2019, 1, 37–47, 2020. 42. Vijayakumaran, C., Muthusenthil, B., Manickavasagam, B., A reliable next generation cyber­ security architecture for industrial internet of things environment. Int. J. Electr. Comput. Eng., 10, 1, 387–395, 2020. 43. Merat, S. and Almuhtadi, W., Artificial intelligence application for improving cybersecurity acquirement. Can. Conf. Electr. Comput. Eng, vol. 2015-June, pp. 1445–1450, 2015. 44. Trifonov, R., Nakov, O., Mladenov, V., Artificial intelligence in cyber threats intelligence. 2018 Int. Conf. Intell. Innov. Comput. Appl. ICONIC 2018, pp. 1–4, 2019. 45. Chan, L., Morgan, I., Gentry, J., Min, D., Ober, D., Survey of AI in Cybersecurity for Information Technology Management. 2019 IEEE Technol. Eng. Manag. Conf., pp. 1–8, 2019. 46. Mathew, A., Airport Cybersecurity & Cyber Resilience Controls. 63–66, 2019. 47. Efthymiopoulos, M.P., A cybersecurity framework for development, defense and innovation at NATO. J. Innov. Entrep., 8, 1, 2019.

Artificial Intelligence and Cybersecurity  441 48. Süzen, A.A., A Risk-Assessment of Cyber Attacks and Defense Strategies in Industry 4.0 Ecosystem. I. J. Comput. Netw. Inf. Secur., February, 1–12, 2020. 49. Reddy, G.N. and Reddy, G.J.U., A Study of Cybersecurity Challenges and Its Emerging Trends on Latest Technologies., February, 2014. [Online]. Available: http://arxiv.org/abs/1402.1842. 50. Cabaj, K., Kotulski, Z., Księżopolski, B., Mazurczyk, W., Cybersecurity: trends, issues, and challenges. Eurasip J. Inf. Secur., 2018, 1, 10–12, 2018. 51. Trifonov, R., Manolov, S., Yoshinov, R., Tsochev, G., Pavlova, G., Artificial Intelligence Methods for Cyber Threats Intelligence. Int. J. Comput., 2, 129–135, 2017. [Online]. Available: https:// www.iaras.org/iaras/filedownloads/ijc/2017/006-0020(2017).pdf.

Index Accelerators, 206 Accelerometers, 253 Actuator, 133, 153, 154, 158, 161, 163 Adaboost algorithm, 317 Advanced driver assistance systems (ADAS), 173 Agriculture, 152, 154, 155, 162 AI, 25 AI in healthcare, 277 Analysis, behavioral, 75 data, 74 predictive, 75 Antidepressant medications, 309 API, 91–93, 96–100, 106–108 Apriori algorithm, 317 Arduino, 15–17, 22 Artificial general intelligence (AGI), 172, 173, 174 Artificial intelligence, 70, 71, 77, 82, 167–180, 252, 261 Artificial intelligence and Internet of Things (IoT), 177 Artificial narrow intelligence (ANI), 172, 173 Artificial neutral network, 339 Artificial super intelligence (ASI), 172, 174, 175 Autoencoders, 200 Automated machine learning (AutoML), 178 Basic security and privacy check list, 374 Basic security concerns for cybersecurity, maintenance, 372–373 precaution, 372 reactions, 373 Behavioral activations, 309 Behavioral analysis, 318–319 Benefits of mental health, early stage detection of chronic disorder, 241 improved drug management, 242

less paperwork and documentation, 241 reduce human error, 241 reduce in treatment cost, 241 reliable results of treatment, 242 speedy medical attention, 242 Big data, 91, 100, 107, 273 4Vs, 274 architecture, 275 variety, 75 velocity, 76 volume, 76 Big data analytics, 134 Big data tools, 279–282 Biometric security, 412–414 Biosensors, 134 Bipolar disorder, 308 Bitcoin, 381, 383 BitPesa, 401 Blockchain, 135, 252 Blockchain in IoT for healthcare, 245 Blockchain technology, analysis, 392 blockchain as a service (BaaS), 401 classification, 381–382 consensus mechanism, 385–386 cryptocurrency and, 384–385 early years of, 389–390 experiment, 394–396 flow chart, 393 history of, 388 introduction, 379–381 literature review, 391 methodology, 392–393 need for, 383–384 proof of stake (PoS), 387–388 proof of work, 386–387 results, 398–400 tools and configuration, 394

443

444  Index Blockchain-based online voting system, 405–407 Brain memory, 318 BS-to-CH feedback trust calculation, 125 Byzantine fault tolerance (BFT), 388 C4.5 algorithm, 317 Cart algorithm, 317 Challenges and issues in IoT, 62 Challenges in IoT-based mental healthcare applications, communications media, 244 computational limits, 243 devices multiplicity, 244 interoperability issues, 243 IoT-based healthcare platforms, 244 memory limitations, 243 network type, 244 quality of service, 245 scalability, 242 security and privacy issues, 243 standardization, 244 trust, 242 CH-to-CH direct trust estimation, 125 CISCO IOx, 329–330 infrastructure as a service, 330 platform as a service, 330 software as a service, 330 Classification of cyber crimes, external attacks, 358 internal attacks, 357 structured attack, 358 unstructured attack, 358 Clinical research, 255 Cloud, 91 Cloud computing, 178, 194, 252 Cloud watcher, 346 Cloud-based secured online voting system, 414–416 Cluster computing, 92, 94 Clustering, 437 CM to CM (direct) trust evaluation scheme, 123–124 CM to CM peer recommendation (indirect) trust estimation (PRx,y(Δt)), 124–125 CNN-U-Net model, 217 Cognitive behavioral therapy (CBT), 309 Cognitive networks, 29 Cognitive radio, 39

Computational intelligence, 287, 301–304 Consensus algorithm, 385–386 Consortium blockchain, 382 Contextual mining, 316 Continuous bag-of-words model, 315 Continuous skip-gram model, 315 Covid-19, 287 Credit card transactions, 384 Cross validation, 72 Cryptocurrency and blockchain technology, 384–385 Cryptographic algorithms, 410–411 Crypto-watermarking approach, 417–418 Customer relationship management (CRM), 175 Cybersecurity and privacy techniques, authentication and authorization, 365–366 cryptography, 366–367 digital signature, 367–368 firewall, 369 installation of antivirus, 367 steganography, 369–370 Cyberattacks, 432 Cybersecurity, 431 Cybersecurity layered stack, 373 Data, imputation, 72 pre-processing, 72 Data analytics, 287, 299, 300, 302 perspective, 81 spatial, 81 streaming, 81 time series, 81 Data collection, 278 Data mining methods, 310–311 Databases and evaluation metrics, 225 CNN implementation details, 226 comparative analysis, 227 evaluation on DRIVE and STARE databases, 227 DataFrames, 91, 92, 96–108 Datasets, 91–99, 103–108 Datasources, 92, 96, 99, 100, 101, 104–108 Deep belief networks, 201 Deep learning, 26, 169, 171, 172, 178, 198, 339 Deep learning model, 217 Delegated proof of stake (DPoS), 387–388 Depression, diagnosis, 309–310

Index  445 treatment, 309–310 types of, 308–309 DHT sensor, 17–18 Diabetes, 254–256 Dielectric, 134 Digital health, 257 Dimensionality reduction, 72 Directed acyclic graphs, 385 Disability and rehabilitation, 254 Distributed ledger technology (DLT), 384–385 Dragonchain, 381–382, 385 Drip irrigation, 160 Electrocardiogram, 264 Electroconvulsive therapy (ECT), 308, 309 Electromyography, 254 Emotion analysis, 318 Encrypted ciphertext, 424 Enhancement, 217 Essential elements of cybersecurity, 370–371 Essential IoT technologies, 193 Ethereum, 381, 390, 399 Facebook, 311, 313 Face-to-face interviews, 309 Facial recognition (FR), 178 Features of IoT, 59 Fingerprint biometric, 417 Fog computing, 325–328 FP-growth algorithm, 317 Future challenges of cybersecurity, 374–375 Future opportunities in IoT, 63 Generative adversarial network, 168, 170 Global positioning system, 254 Google predictive search, 175 Group interviews, 309 Gujarati joint characters, 179, 180 Gyroscopes, 253 Hadoop, 92–94, 108 Happiness, 287, 293–298 Hashgraph, 385 Healthcare, 251, 252, 256, 257 Healthcare data, 276 Heterogeneity, 4, 14, 22 Historical background and evolution of cyber crime, 354–355 Hive tables, 92, 96, 98, 99, 105

Homomorphic encryption, 414–416 Humidity, 155, 157, 158, 163, 165 Hybrid blockchain, 382 Hybrid sensor network, 139 Hyper parameter tuning, 72 Hyperledger, 381 IIoT, 81 Image recognition, 204 Industry 4.0, 283 Instagram, 311 Intelligent machine, 168 Internet data, 310 Internet of Things, 71, 80, 133, 186, 252 Interpersonal psychotherapy (IPT), 309 Introduction to cybersecurity, application security, 356 information security, 356 network security, 357 recovery from failure or disaster, 356–357 Investopedia, 401 IoT, 153, 154, 158, 163 IoT application, 3, 7, 15, 19–20 IoT architecture, 5, 61 IoT communication protocol, bluetooth, 3, 12–13 IEEE 802.15.4, 12–14 NFC, 12–13 wireless-hart, 12–14 Zigbee, 12–13 Z-wave, 12, 14 IoT connectivity technologies, 6LoPAN, 11–12 AMQP, 11 CoAP, 10–11 MQTT, 11 RFID, 7–8, 10–11, 13 XMPP, 10–11 IoT in mental health, 238 IoT networking, 8, 10, 13 IoT sensor devices, 59 IOTA, 385 Iris recognition, 418–420 ITS, 145 Java EE application server, 422–423 JSON, 92, 98–102, 105, 106 K-means, 34, 317 K-nearest neighbors, 317

446  Index KNN, 33 KPI, 29 Litecoin, 399 M2M communication, 6–7, 10–11, 14 Machine learning, 30, 36, 70, 71, 169–172, 177, 252, 262, 338 Bayesian network, 340 decision tree, 340 reinforcement learning, 340 Machine learning models, 315 Magnetometers, 253 Major depression disorder, 308 Malicious node detection, 127–128 Mantra, 287, 289–291, 293, 295–299, 301 MapReduce, 92, 94, 95, 108 Mastercard, 384, 396, 399–400 Medical imaging, 264, 268 Mental healthcare applications and services based on IoT, 238 Methodology, 223 architecture of Stride U-Net, 223 loss function, 225 Microcontroller, 254 Middleware, 194 ML, 30 Mobile robotics, 209 Model selection, 72 Moisture, 153, 154, 160, 162, 163, 165 Mood swing problem, 308, 318 Naïve Bayes algorithm, 317 National identification (NID) numbers, 413, 420–422 Natural language processing (NLP), 170, 171, 172, 175, 311, 312, 316–317 Network compression, 205 Network security, 433 Network topology and assumptions, 122 Neural network, 168, 171, 180 Neural network models, 315 New cache, 346 NLP toolkit (NLTK), 313 Om, 287, 305 Online conversations, 318 Open flow, 38 Optical character recognition, 413

Paper rank algorithm, 317t Parallelism, 71 Parquet files, 92, 99, 106 Patient data, 276 Patient diagnosis, 278 Patient monitoring, 282 wearable devices, 282 Pattern recognition, 26 Persistent depressive disorder, 308 Pixel encoding schemes, 411 Precision health, 251, 252, 255, 257 Principal component analysis (PCA), 217 Privacy and security, 190 Privacy-preserving aggregation, 345 Private blockchain, 382 Proof of stake (PoS), 387–388 Proof of work (PoW), 380, 386–387 Proposed trust model, 122–125 Psychotherapy, 308 Psychotic depression, 309 Public blockchain, 381 Questionnaires, 309 R3 Corda, 381 Raspberry pi, 15, 17–18, 22 RDD, 91–93, 95, 96, 98–102, 104, 108 Reasons behind cyber crime, business analysis and decision making, 359 for making fun, 359 gaining financial growth and reputation, 359 making money, 359 revenge, 359 to recognize, 359 Reference architecture, 328 Reinforcement learning, 35, 171, 262 Repetitive transcranial magnetic stimulation (rTMS), 308 Resilience, 139 Resource management, 328–329 Retinal color fundus images, 217 RF module, 153–165 RSA encryption, 411 Seasonal effective disorder (SAD), 308 Secret sharing–based authentication, 425 Secured online voting systems, blockchain-based online voting system, 405–409

Index  447 comparison of existing systems, 427–429 literature review, 410–425 Security information and event management (SIEM), 176 Segmented retinal vessel image, 217 Sensing, 41 Sensitivity boosting, 217 Sensor, 154–156, 160, 162–164 Sensors, 133 Sentiment analysis, 316–319 algorithms and approaches, 318 Severity analysis, 126–127 SHA-256 hashing, 407–408 SIRI, 168, 170, 172 Smart agriculture, 143, 144 Smart city, 136, 137, 138 Smart device/sensor layer, 187 Smart farm, 3, 5, 19–21 Smart healthcare, 139, 140, 141, 142 Social media, 311–312, 316 Social network sites, 310–312 Software defined networks, 36 Software malfunction, 177 SON, 27 Spark SQL, 92, 93, 97–108 SQLContext, 92, 100 Stablecoin, 399 STEALTHMEM, 346 Steganographic systems, 410–411 Steganography, 412–414 Subscriber identification module (SIM) card, 420–422 Supervised learning, 32, 171, 262 Support vector machine, 33, 317 Symptom extraction tool, 312–316 Technical interoperability, 189 Text mining, 310–311 Text mining algorithms, 317 Thinking machines, 168 Tiny vessel detections, 217 Tooth sensors, 253 Transducer, 9, 12

Turing test, 168 Twitter, 313 Unsupervised learning, 33, 171, 262 Unsupervised learning algorithm, 315, 316 Various types of cyber attacks in information security, system-based attacks in information security, 364–365 web-based attacks in information security, 362–363 Various types of cyber crime, crime related to privacy of software and network resources, 360 cyber stalking, 360 cyber terrorism, 360 forgery, 360 malfunction, 361 phishing, vishing, and smishing, 360–361 server hacking, 361 sexual harassment or child pornography, 360 spamming, cross site scripting, and web jacking, 361 spreading virus, 361 Virtual assistance, 173 Virtual private network, 346 Virtual reality, 252 Visa, 396, 399–400 Visa exchange, 384 Visual cryptography, 411 Visual cryptography algorithm, 425 Walmart, 401 WBAN, 141 Wearables, 251, 252, 255–257 Word clustering, 315–316 Word embedding, 314–316 Word2Vec, 315 Yajna, 287, 289–299, 301, 303–305

 

-

WILEY END USER LICENSE AGREEMENT Go to www.wiley.com/go/eula to access Wiley’s ebook EULA.