Recent Advances in Computational Intelligence [1st ed.] 978-3-030-12499-1, 978-3-030-12500-4

Table of contents :
Front Matter ....Pages i-ix
Front Matter ....Pages 1-1
Smart Inventory Management Using Electronic Sensor Based Computational Intelligence (Kunal Singh)....Pages 3-25
Computational Intelligence for Data Analytics (S. Anupama Kumar, M. N. Vijayalakshmi, T. L. Divya, K. N. Subramanya)....Pages 27-43
Design and Analysis of an Enhanced Multifactor Authentication Through a Covert Approach (Raman Kumar, Uffe Kock Wiil)....Pages 45-61
Review on Current Trends of Deep Learning (Stuti Mehla, Anjali Chaudhary, Raman Kumar)....Pages 63-70
Hybrid Homomorphic Encryption Scheme for Secure Cloud Data Storage (Bijeta Seth, Surjeet Dalal, Raman Kumar)....Pages 71-92
A Gamification Framework for Redesigning the Learning Environment ( Rishipal, Sweta Saraff, Raman Kumar)....Pages 93-105
A Variant of Secret Sharing Protected with Poly-1305 (Shakti Arora, Surjeet Dalal, Raman Kumar)....Pages 107-119
Securing Bioinformatics Cloud for Big Data: Budding Buzzword or a Glance of the Future (Bijeta Seth, Surjeet Dalal, Raman Kumar)....Pages 121-147
Enhancing Gadgets for Blinds Through Scale Invariant Feature Transform (Raman Kumar, Uffe Kock Wiil)....Pages 149-159
Front Matter ....Pages 161-161
A Study to Renovate Image Data Using Data Analytics Methodologies (A. Prema)....Pages 163-171
The Use of Modern Technology in Smart Waste Management and Recycling: Artificial Intelligence and Machine Learning (Praveen Kumar Gupta, Vidhya Shree, Lingayya Hiremath, Sindhu Rajendran)....Pages 173-188
Video Streaming Communication over VANET (Gurpreet Singh)....Pages 189-197
An Update on Effective Patient Monitoring and Diagnosis of Diseases in the Current Scenario (Sindhu Rajendran, Meghamadhuri Vakil, Praveen Kumar Gupta, Lingayya Hiremath, S. Narendra Kumar, Ajeet Kumar Srivastava et al.)....Pages 199-214
Automation of Assets Estimation for Accurate Tax Optimization Through Web Based Spatial Data Mining Techniques (V. Pattabiraman, R. Parvathi)....Pages 215-232
Computationally Efficient and Secure HVS Based Composite Fingerprinting Scheme Using SHS and ECDSA (Vineet Mehan)....Pages 233-254
Autonomous Agents—Beginnings, Innings and Where We Are Headed (Minal Moharir, Ananthakrishnan Soumya Mahalakshmi, Gupta Praveen Kumar)....Pages 255-261
Modern Methods for Signal Analysis and Its Applications (K. P. Soman, S. Sachin Kumar, N. Mohan, P. Poornachandran)....Pages 263-290
Word Sense Disambiguation of Malayalam Nouns (S. N. Mohan Raj, S. Sachin Kumar, Sindhu Rajendran, K. P. Soman)....Pages 291-314
A Comparative Analysis of Machine Comprehension Using Deep Learning Models in Code-Mixed Hindi Language (Sujith Viswanathan, M. Anand Kumar, K. P. Soman)....Pages 315-339
Embedding Linguistic Features in Word Embedding for Preposition Sense Disambiguation in English—Malayalam Machine Translation Context (B. Premjith, K. P. Soman, M. Anand Kumar, D. Jyothi Ratnam)....Pages 341-370
Translation Equivalence for English Periphrastic Causative Constructions into Hindi in the Context of English to Hindi Machine Translation System (D. Jyothi Ratnam, K. P. Soman, T. K. Biji Mol, M. G. Priya)....Pages 371-392

Citation preview

Studies in Computational Intelligence 823

Raman Kumar Uffe Kock Wiil   Editors

Recent Advances in Computational Intelligence

Studies in Computational Intelligence Volume 823

Series Editor Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland

The series “Studies in Computational Intelligence” (SCI) publishes new developments and advances in the various areas of computational intelligence—quickly and with a high quality. The intent is to cover the theory, applications, and design methods of computational intelligence, as embedded in the fields of engineering, computer science, physics and life sciences, as well as the methodologies behind them. The series contains monographs, lecture notes and edited volumes in computational intelligence spanning the areas of neural networks, connectionist systems, genetic algorithms, evolutionary computation, artificial intelligence, cellular automata, self-organizing systems, soft computing, fuzzy systems, and hybrid intelligent systems. Of particular value to both the contributors and the readership are the short publication timeframe and the world-wide distribution, which enable both wide and rapid dissemination of research output. The books of this series are submitted to indexing to Web of Science, EI-Compendex, DBLP, SCOPUS, Google Scholar and Springerlink.

More information about this series at http://www.springer.com/series/7092

Raman Kumar Uffe Kock Wiil •

Editors

Recent Advances in Computational Intelligence

123

Editors Raman Kumar Department of Computer Science and Engineering I. K. Gujral Punjab Technical University Kapurthala, Punjab, India

Uffe Kock Wiil The Maersk Mc-Kinney Moller Institute University of Southern Denmark Odense, Denmark

ISSN 1860-949X ISSN 1860-9503 (electronic) Studies in Computational Intelligence ISBN 978-3-030-12499-1 ISBN 978-3-030-12500-4 (eBook) https://doi.org/10.1007/978-3-030-12500-4 Library of Congress Control Number: 2019930567 © Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

In recent years, there has been an increasing focus on computational intelligence including the theories, methodologies and techniques underlying this evolving field as well as its potential use in various domains across the entire spectrum of sciences (natural science, health science, engineering, social science and humanities) and in various types of businesses. Computational intelligence has yet to achieve its full potential and is projected to play a major role in the development of successful future intelligent systems. Recent Advances in Computational Intelligence is an edited book that contains contributions from various experienced professionals ranging from the foundations of computational intelligence to examples of its use across multiple disciplines. The book has enabled scientists, scholars, engineers, professionals, policy-makers, government and non-government organizations to share new developments in theory, analytical and numerical simulation and modelling, experimentation, demonstration, advanced deployment and case studies, results of laboratory or field operational tests and ongoing developments with relevance to advances in computational intelligence. The book is divided into two parts comprising nine and twelve chapters, respectively. The first part includes chapters that primarily focus on the foundations of computational intelligence. The chapters deal with the following topics: (1) smart inventory management using electronic sensor-based computational intelligence; (2) computational intelligence for data analytics; (3) design and analysis of an enhanced multifactor authentication through a covert approach; (4) review on current trends of deep learning; (5) hybrid homomorphic encryption scheme for secure cloud data storage; (6) a gamification framework for redesigning the learning environment; (7) a variant of secret sharing protected with Poly 1305; (8) securing bioinformatics cloud for big data: budding buzzword or a glance of the future; and (9) enhancing gadgets for blinds through scale-invariant feature transform. The second part includes chapters that primarily demonstrate, through detailed studies and cases, how computational intelligence can be used (and is already in use). The following studies and cases are included: (10) a study to renovate image v

vi

Preface

data using data analytics methodologies; (11) the use of modern technology in smart waste management and recycling: artificial intelligence and machine learning; (12) video streaming communication over VANET; (13) an update on effective patient monitoring and diagnosis of diseases in the current scenario; (14) automation of asset estimation for accurate tax optimization through Web-based spatial data mining techniques; (15) computationally efficient and secure HVS-based composite fingerprinting scheme using SHS and ECDSA; (16) autonomous agents —beginnings, innings and where we are headed; (17) modern methods for signal analysis and its applications; (18) word sense disambiguation of Malayalam nouns; (19) a comparative analysis of machine comprehension using deep learning models in code-mixed Hindi language; (20) embedding linguistic features in word embedding for preposition sense disambiguation in English–Malayalam machine translation context; and (21) translation equivalence for English periphrastic causative constructions into Hindi in the context of English to Hindi machine translation system. This book can benefit researchers, advanced students as well as practitioners. The collection of papers in the book can inspire future researchers—in particular, researchers interested in interdisciplinary research. The rich interdisciplinary contents of the book can be of interest to faculty, research communities, and researchers and practitioners from diverse disciplines who aspire to create new and innovative research initiatives and applications. The book aims to inspire researchers and practitioners from different research backgrounds regarding new research directions and application domains within computational intelligence. We wish to thank all the people that contributed to this edited book. We wish to thank the authors for their insightful contributions, the reviewers for their suggestions that ensured the quality of the individual parts and, last but not least, the Springer Team for their continuous support throughout the project. With this joint effort, this book would not have been possible. Kapurthala, Punjab, India Odense, Denmark

Dr. Raman Kumar Prof. Uffe Kock Wiil

Contents

Part I

Foundations of Computational Intelligence

Smart Inventory Management Using Electronic Sensor Based Computational Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kunal Singh Computational Intelligence for Data Analytics . . . . . . . . . . . . . . . . . . . S. Anupama Kumar, M. N. Vijayalakshmi, T. L. Divya and K. N. Subramanya Design and Analysis of an Enhanced Multifactor Authentication Through a Covert Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Raman Kumar and Uffe Kock Wiil Review on Current Trends of Deep Learning . . . . . . . . . . . . . . . . . . . . Stuti Mehla, Anjali Chaudhary and Raman Kumar

3 27

45 63

Hybrid Homomorphic Encryption Scheme for Secure Cloud Data Storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bijeta Seth, Surjeet Dalal and Raman Kumar

71

A Gamification Framework for Redesigning the Learning Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rishipal, Sweta Saraff and Raman Kumar

93

A Variant of Secret Sharing Protected with Poly-1305 . . . . . . . . . . . . . Shakti Arora, Surjeet Dalal and Raman Kumar Securing Bioinformatics Cloud for Big Data: Budding Buzzword or a Glance of the Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bijeta Seth, Surjeet Dalal and Raman Kumar

107

121

vii

viii

Contents

Enhancing Gadgets for Blinds Through Scale Invariant Feature Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Raman Kumar and Uffe Kock Wiil Part II

149

Applications of Computational Intelligence

A Study to Renovate Image Data Using Data Analytics Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Prema The Use of Modern Technology in Smart Waste Management and Recycling: Artificial Intelligence and Machine Learning . . . . . . . . Praveen Kumar Gupta, Vidhya Shree, Lingayya Hiremath and Sindhu Rajendran Video Streaming Communication over VANET . . . . . . . . . . . . . . . . . . Gurpreet Singh An Update on Effective Patient Monitoring and Diagnosis of Diseases in the Current Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sindhu Rajendran, Meghamadhuri Vakil, Praveen Kumar Gupta, Lingayya Hiremath, S. Narendra Kumar, Ajeet Kumar Srivastava and Vidhya Shree

163

173

189

199

Automation of Assets Estimation for Accurate Tax Optimization Through Web Based Spatial Data Mining Techniques . . . . . . . . . . . . . V. Pattabiraman and R. Parvathi

215

Computationally Efficient and Secure HVS Based Composite Fingerprinting Scheme Using SHS and ECDSA . . . . . . . . . . . . . . . . . . Vineet Mehan

233

Autonomous Agents—Beginnings, Innings and Where We Are Headed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Minal Moharir, Ananthakrishnan Soumya Mahalakshmi and Gupta Praveen Kumar

255

Modern Methods for Signal Analysis and Its Applications . . . . . . . . . . K. P. Soman, S. Sachin Kumar, N. Mohan and P. Poornachandran

263

Word Sense Disambiguation of Malayalam Nouns . . . . . . . . . . . . . . . . S. N. Mohan Raj, S. Sachin Kumar, Sindhu Rajendran and K. P. Soman

291

A Comparative Analysis of Machine Comprehension Using Deep Learning Models in Code-Mixed Hindi Language . . . . . . . . . . . . Sujith Viswanathan, M. Anand Kumar and K. P. Soman

315

Contents

Embedding Linguistic Features in Word Embedding for Preposition Sense Disambiguation in English—Malayalam Machine Translation Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Premjith, K. P. Soman, M. Anand Kumar and D. Jyothi Ratnam Translation Equivalence for English Periphrastic Causative Constructions into Hindi in the Context of English to Hindi Machine Translation System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Jyothi Ratnam, K. P. Soman, T. K. Biji Mol and M. G. Priya

ix

341

371

Part I

Foundations of Computational Intelligence

Smart Inventory Management Using Electronic Sensor Based Computational Intelligence Kunal Singh

Abstract With the advent of computational technology and digitalization of data, newer methods have been developed for inventory management using computational intelligence. Electronic Sensor based technologies combined with this intelligence can create a remarkable combination of inventory management, which can not only keep a thorough track of the goods required, but the system is also very easy to manage, as a consequence of which these sensors based smart inventory management system become very efficient in terms of preventing any mismanagement of goods. In this chapter, various such smart systems are discussed in detail by the author. The drawbacks of traditional inventory management system and the effect of computational intelligence on improvisation of these systems have been discussed in detail. The principle of operation, Design and Applications of RFID (Radio Frequency Identification), Load Cells and Ultrasonic sensor based inventory management systems are discussed in detail. Also, some recent advancements in inventory management like three-dimensional scanning system along with some other sensor based smart inventory systems proposed by the author have been discussed in detail. Methods on improving the portability and efficiency in terms of ease of use of such systems have been discussed in detail, which can be achieved using Bluetooth or wireless radio communication. Keywords Inventory · Sensor · RFID · Load cell · Ultrasonic sensor

1 Introduction Inventory management is an important characteristic of a good supervised business. It supervises the availability of goods which need to flow from manufactures to the warehouses and ultimately to the consumers. Supervision of inventory not only helps K. Singh (B) Mechanical Engineering Department, Maharaja Agrasen University, Baddi 174103, HP, India e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_1

3

4

K. Singh

the business to cut unwanted costs of travel and transportation but also helps to meet customer demands at all times. As a consequence of such supervision, a company may attain high profits and a good customer review which ultimately adds to the company’s good reputation in the market. The traditional method for inventory management, without any computational intelligence or sensor technology may help in forecasting customer demands by analyzing the required goods based on previous sale experience. However, a smart inventory management [1] is not just all about forecasting the demands, but also monitoring real time requirement. For example, a store which sell goods to their customers, based on this traditional approach, purchased goods may reach warehouses, but may not reach the stores on time. So, even though the purchase is done on time based on forecasting, the sales may suffer due to mismanagement of goods. This loss may be attributed to the miscommunication between warehouse and store. It may also happen that due to human error wrong material or unwanted goods may be transported to the store, which is again a loss for the company as well as the customer. Such problems can be prevented by adopting a system which not only monitors the forecasting of goods but also the real-time analysis of availability of goods for the customers. Traditional systems are incapable of solving these problems as it requires a simultaneous monitoring of demand and availability of goods to the customers. However, with the advancements in computational technologies, a software based system is much more efficient in eliminating these problems. Such systems use computational intelligence, and utilize their logical decision-making capabilities to communicate between warehouses, manufacturers and customer themselves. Such smart inventory management systems can monitor and record real-time requirements of the customers and may help establishing good customer relationships. In order to fully utilize the capability of such systems, advanced sensor based technologies must be coupled with these systems, as these sensors are fast and much more accurate than humans in estimating the quality or quantity of the goods required. In this chapter, various such sensor based inventory management systems are discussed in detail which may provide lighting fast real-time information of the goods to the computer program, which can estimate not only the current requirement based on this data, but also forecast the requirement based on the collection of such previous data lists. Based on the sensor data, computer program may realize micro requirements, i.e. requirement of goods on peak hours of a day, and may effectively communicate warehouses for the same. With the advancements in automation and engineering, sensor technology is on its way to become smaller and smarter. The efficiency of sensors in terms of power consumption and digital resolution is increasing remarkably, as these sensors are being designed to integrate efficiently with intelligent programming software and hardware systems which require low power consumption components. Due to reduction in size along with new and efficient engineering techniques of manufacturing these sensors, the cost of sensors is also reducing which plays an important role in their utilization with the smart inventory systems. Due to small size and low cost the sensors are becoming an integral part of the product themselves without affecting much the overall cost of the product. It has been observed through studies that the demand of the current era is customer

Smart Inventory Management Using Electronic Sensor …

5

satisfaction. The customers are willing to pay an extra amount for good quality and availability of goods. Such conditions can only be satisfied by the adoption of smart sensor based inventory systems.

2 Sensor Based Smart Inventory System Before we start our discussion on techniques that may be employed to create such smart inventory systems, it may be appropriate to define the meaning of the word “Sensor” or more conveniently a “Digital sensor”. A “Digital sensor” is a device that consumes minimal amount of energy from the system but tries to sense the exact state of the system in digital form. This state can be temperature, pressure, quantity, quality, volume etc. Most of the digital sensors have an inbuilt microprocessor on the circuit board which convert the analog input data to digital output form. It is important to note that the real world is analog in nature. But, it is up to the user to define the resolution of this analog world based on the requirement. The analog input data such as change in temperature, pressure or volume of a system may affect the resistance, capacitance or inductance of a sensor. This change in electrical quantities can be captured in digital form using microprocessors or ADC [2] (Analog to digital converters). Such data is very useful and is appropriate for analysis through a computer program. Enlisted below are some of the smart sensor based technologies that are being developed in order to integrate them with computer intelligence driven inventory management systems. • RFID (Radio Frequency Identification) based inventory management. • Load Cell based inventory management • Ultrasonic sensor based inventory management. The principle of operation of the above-mentioned inventory systems have been discussed in detail in this chapter. Also, an appropriate design and circuitry of these sensors for the implementation in an inventory system have been discussed in detail. It can be observed that the application of such system is specific to the nature of goods. For example, RFID [3] based system is suitable for packaged goods, while Ultrasonic sensor based system is suitable only for liquefied goods, on the other hand Load cell based system can be implemented both for packaged or unpackaged goods or liquefied goods. Such applications, determining the area of their implementation have been discussed in detail.

2.1 RFID (Radio Frequency Identification) Inventory System This technology is the most preferred method for smart inventory system, as it is very easy to implement and require no previous estimation or calculation of the size or volume of components required in the inventory. The basic principle of RFID is

6

K. Singh

Fig. 1 The distance up to which RFID system can work depends on the frequency range

the identification of radio frequency which is achieved through load modulation [4] process for near field coupling or backscattering [5] process for far field coupling. RFID system consists of two components, one is RFID reader and other is RFID tag. The purpose of RFID system is to generate and maintain a magnetic field around its coil using an alternating current. The frequency of this alternating current can be broadly divided into three frequency range, which are low frequency (125–134 kHz), High frequency (13.56 MHz) and Ultra high frequency (860–960 MHz). In general, higher the frequency more is the range of operation of this system as shown in Fig. 1. RFID tag is the component of this system which is detected by the reader. Therefore, it can be attached to the goods which need to be detected by the RFID reader. It can be noticed that as this system works on radio frequency, the RFID reader and tags must not necessary be in line of sight with each other as is the case with bar code scanners. In general, there are three types of RFID tags which are categorized based on their power requirement. These are Active tags, Passive tags [6] and Semi Passive tags [7]. In case of active tags, the tags have their own power supply to transmit the feedback signal to the reader. Even though the signal strength of this tag is strong, this type of tag is not suitable for inventory systems as the battery of tag will drain with time which will render the tag useless. Semi Passive tags on the other hand have their own power supply to run the internal circuitry, but this power is not used to transmit the feedback signal back to the reader, in fact they use the energy of the radio frequency of the reader to generate a feedback signal via the process of inductive coupling or electromagnetic induction, which is in fact the same principle on which a transformer works. The transponder of the tag gets coupled with the magnetic field generated by the coil of the reader and generates voltage for the feedback signal. The advantage of such tags is that their battery may last longer as it is only being used for running internal circuit of tag and the feedback signal is strong as the power generated by the transponder is only being utilized for this feedback signal. Such tags are more suitable then active tags due to long battery life and strong signal strength. The next category is of Passive tags, which rely totally on the power generated by the transponder to run the circuit within the tag as well as generation of feedback signal. This type of tag is best suited for a smart inventory system as they can be weather sealed within water proof body, and require no power supply, which makes them inexpensive. In this tag, a part of the power supply received by the transponder is rectified and transmitted to the circuitry while other part is used

Smart Inventory Management Using Electronic Sensor …

7

Fig. 2 Components of a RFID Reader

Fig. 3 Near field coupling between RFID reader and RFID tag working with low frequency or high frequencies

for signal generation. Before we discuss the circuitry of the RFID systems, let us look at the basic components of RFID reader as shown in Fig. 2. It may be noted that there are two separate coils for generating carrier frequency and receiving the feedback signal. Figure 3 depicts the NFC coupling between RIFD reader and tag which is valid for Low frequency and High frequency RFID systems. Note that the tag used here is purely passive in nature. Rectifier is used to convert a part of the alternating voltage to DC voltage for the microcontroller. A transistor whose collector and emitter are attached to both the ends of the transponder coil is closed and open based on the data or code generated by the microcontroller pin, which is connected to the transistor base. This process causes the variation of load in the transponder coil and is called load modulation. Due to this load modulation, the RFID generator coil of the reader suffers a variation in power consumption which can be detected by the RFID receiver coil. It may be noted that this unique variation of load modulation can provide a unique identification to a tag and thus can identify the product on which it is attached or embedded within. It may be noted that the clock generated for the microcontroller inside the tag is driven by the carrier frequency itself and no external clock source may be required. The hardware of the far field RFID systems is similar to as depicted in Fig. 3. The difference is that instead of electromagnetic coupling between the two coils the coils of the reader and the tag act as antennas, which transmits and receives

8

K. Singh

Fig. 4 On the left is an inexpensive RFID reader with some passive tags on the right

Fig. 5 Possible algorithm for application for RFID in a smart inventory system

radio frequencies. The generator coil of the reader sends high frequency radio waves which are received by the transponder. Due to this fluctuating radio energy, current is induced in the transponder, which generate its own magnetic field in the form of radio energy and is also called back scattered energy [5]. Again, a little short circuit using transistor will affect this radio energy generated by the transponder and can be coded in the form of unique identification number. Figure 4 shows an inexpensive RFID reader along with some passive RFID tags that can be utilized for smart inventory systems. Figure 5 represents an algorithm that can employed for smart inventory management using RFID system through computational intelligence. From Fig. 5 it can be observed that in this algorithm RFID reader transmits signals continuously when applied to an inventory system, until it detects the required

Smart Inventory Management Using Electronic Sensor …

9

modulated signal. If the signal received is correct then the counter of product is incremented by one. This algorithm is applicable when product is transported to the inventory or on the shelves of the store. The reader can be attached on the entry door of the inventory room or the store. As the product pass through the entry door the product will automatically get register in the inventory by a computer which runs a program based on this algorithm. However, for the removal of product a separate exit door must be used which runs similar algorithm except that now the counter must decrement to register a decrease in the amount of the product.

2.2 Load Cell Inventory System It’s quite remarkable how computational intelligence have rapidly evolved in the recent years. The tasks that seem to be tedious or impossible in long run can be achieved easily through a simple computer programs. “Load cell” [8] based inventory system is a remarkable example, expressing beautifully the intelligent behavior of computer programs. In this type of inventory system any change in the weight of the goods is precisely calculated by a computer program and logical decisions are made whether the inventory require more stock or not. In this system load Cells which are basically metal spring elements are used which consists of “Strain gauges” [8]. It may be noted that the resistance of a “Strain gauge” change when a load or weight is applied on it. This change in resistance can be sensed by utilizing the principle of Wheatstone bridge, which estimates the null voltage change between two potential dividers connected in parallel. For this purpose, the strain gauges can be arranged in quarter, half or full Wheatstone bridge configuration. It has been observed that full Wheatstone bridge configuration is very sensitive to any change in resistance and therefore mostly used in load cells for better sensitivity. In this mode, all the four resistances of Wheatstone bridge are variable in nature, which means we need 4 strain gauges for this type of configuration. Figure 6 depicts the arrangement of strain gauges for this configuration in the form of a circuit diagram. It may be noted that the voltage “V” as shown in Fig. 6, will be zero when the ratio R1/R3  R2/R4, where, R1, R2, R3 and R4 are the respective strain gauges in a load cell. The gauges are so arranged within a load cell that resistance R1 and R4 will experience tensile stress while R2 and R3 bear compressive stress, as with this condition the sensitivity of the bridge is maximum. Figure 7 shows a CAD model of a typical load cell. From the top isometric view and bottom isometric view we can observe how these strain gauges are placed. It may be noted that the placement of strain gauges is not random, but they are placed where stresses within load cell are maximum. As these strain gauges are pasted onto the load cell, they will experience similar stresses faced within the load cell. Figure 8 shows the results of the static stress analysis performed on the CAD model of a load cell made of aluminum. In this analysis, a point load of 100N is applied on the right while the left side of load cell is fixed. The displacement of load cell is highly exaggerated to visually analyze the stresses and to observe how tensile

10

K. Singh

Fig. 6 Full bridge strain circuit. Image downloaded from https://www.allaboutcircuits.com/ textbook/direct-current/chpt-9/strain-gauges/ on 21st May, 2018

Fig. 7 Placement of strain gauges within a load cell

Smart Inventory Management Using Electronic Sensor …

11

Fig. 8 Static stress analysis of load cell depicting nature of tensile and compressive stresses

and compressive stresses are generated within load cell at particular locations only. Also, as the stresses at these locations are almost equal (Range: 20.44–21.89 MPa), it can be interpreted that the change in resistance within the four strain gauges will be equal if placed at these locations. Therefore, the position of strain gauges to obtain optimal sensitivity can be confirmed through this analysis. As we now understand how we can convert the change in resistance of strain gauges to a much usable form such as voltage difference and also this voltage is analog in nature, we need to amplify this difference using operational amplifiers and convert it into digital form before processing it through a computer program, for which we need an analog to digital converter (ADC). An example of such converter is “Dual channel HX711 pressure sensor module” as shown in Fig. 9, which is a 24-bit ADC. This module has an inbuilt amplifier which can provide a gain of 128. Such modules enhance the sensitivity and precision of measurement of load cells. The precision of such modules can be estimated from a simple calculation as follows. Suppose, the input to the load cell is 5 V, then the analog output of load cell will vary from 0 to 5 V. The 24-bit output of ADC means that the minimum resolution we can obtain using 5 V as input to load cell is 5/224 , which is 0.298 µA. Such resolution is sufficient enough to measure the weight of objects within decimal values of grams. The digital output from this module can be fed directly to the inputs of a microcontroller or PLC (Programmable logic controller), which can process the information based on the requirement of inventory system using a computer program. For example, the load cell can be attached beneath the base of a storage crate, the output of this load cell is given to the HX711 pressure sensor module which outputs the digital values to PLC or microcontroller. Microcontrollers or a computer can be programmed using an algorithm as shown in Fig. 10, which calculates the number of items in the crate based on their weight. The condition for such system to work effectively is that the weight of each of the goods inside the crate should be same, which is practically possible for standardized packaged products.

12

K. Singh

Fig. 9 Dual channel HX711 pressure sensor with 24-bit precision ADC module

It can be observed, that in this algorithm the number of items are calculated by dividing total weight with the weight of one item. Suppose, this division gives a float value of one decimal point, then we need to multiply it by 10 to get an integer value. This value is stored in an integer “Count”. Note that the type cast method “(int)” is utilized to convert the float values to an integer. Now the variable “Count” has an integer value which we need to process to estimate the actual number of items in the crate. For this purpose, we can find the remainder of “Count” when divided by 10 using modulo “%” operator, if the remainder is greater than 5, then we can estimate that one of the item have quantity little bit less than the actual standard value. Therefore, to take that into account we divide integer “Count” by 10 to eliminate the last decimal value and add 1 to complete its count in whole numbers. Using this method, a greater than 5 decimal value is replaced with a complete number. Similarly, a less than 0.5 decimal value is totally eliminated by subtracting it from the “Count”, suggesting that few items might have greater weight than standard that might increase the overall weight a little, but it doesn’t mean that there is an extra item. This simple system of interpretation is only valid if the items in the crate are expected to have a precise standard weight. In the real world, a little bit variation in weight is expected and can logically be tackled as explained in the above algorithm. It may be noted that above algorithm is just one solution of obtaining the number of items from the total weight. For example, the actual computer program can be made much more complex, such that in order to interpret accurately the number of items, it may analyze previous data of weights. This complex program may have data logging ability to record previous weight and analyze the probable and most prominent value of number of items in that crate at that particular time. Such algorithm might well tackle the uncertainties in the individual weight of items, and can keep track of inventory with respect to time. A network of such crates can calculate the overall requirement of goods at a particular time. The advantage of such system is that it can be applied on the shelves of the market itself where customers can interact with these smart crates themselves. Figure 11 shows the basic circuit diagram of a load

Smart Inventory Management Using Electronic Sensor …

13

Fig. 10 Alogorithm for inventory estimation using load cell

cell connected with HX711 pressure sensor and a microcontroller (Arduino Nano, Atmega 328P). A program based on the algorithm as mentioned in Fig. 10 can be employed to run this system. Note that the circuit is simplified, so that an overall picture of such system can be analyzed easily by the reader. The actual system might require more components like indicators, switches, relays etc. However, the heart of the system is the computer program itself which will run this system. The DT and SCK pins of the output pins of HX711 module is connected to two digital pins of a microcontroller, in this example they are connected to Pin 2 and 3.

14

K. Singh

Fig. 11 Wiring between load cell, HX711 module and microcontroller for a smart inventory system

2.3 Ultrasonic Sensor Based Inventory System This technology is based on the principle of distance calculation by estimation of the time of propagation of an ultrasonic sound wave [9]. Such precise calculation can only be estimated through a computer program. The inventory system based on this technology is particularly suitable for goods stored in liquid forms. Using this technology, the level of liquid can be accurately calculated in a storage container. Any drop in the level of liquid in the storage containers can be precisely calculated and recorded within fraction of a second using a computer program. The main advantage of using ultrasonic sensor for such inventory system is that it doesn’t interfere physically with the liquid stored. The sensor purely interacts with the liquid using ultrasonic soundwaves. It may be noted that as there is no physical contact of sensor with the liquid, the sensor part of this inventory system becomes modular in nature. The size and specifications of this ultrasonic sensor module is independent of the size of the container and can easily be attached to almost any storage container. An ultrasonic sensor module consists of two parts, one is ultrasonic wave transmitter and other is sound receiver. The purpose of transmitter is to generate ultrasonic sound waves which strike the surface of liquid and reflects back to the receiver. The amount of time taken by the soundwaves to travel from transmitter to receiver is recorded and based on the speed of sound which is approximately 343 m/s, a computer program calculates the distance between the sensor module and the surface of water. This distance is an estimation of the liquid left in the container. Using this system an alarm can be activated to alert low levels of liquid or a valve can be automatically activated to maintain the level of liquid in the container. Figure 12 shows an ultrasonic sensor module (HC-SR04) which is inexpensive but is very effective in estimating precise level of liquid. As shown in Fig. 12 there are four pins in this sensor, VCC and GND pins are used for 5-V DC input, while TRIG pin is used to activate the transmitter. The ECHO

Smart Inventory Management Using Electronic Sensor …

15

Fig. 12 Ultrasonic sensor (HC-SR04) for smart inventory system

pin is activated when the sound receiver senses the reflected signal. Time between sound wave transmission and reception is generated in microseconds by this module which can be accessed via the ECHO pin. Figure 13 depicts the working of this sensor module. For example, if the trigger pin is activated for 10 µS then transmitter is activated for this time period and generates a soundwave which is approximately 8 cycle sonic wave. This sound travels towards the liquid in the container and is reflected back towards the receiver. Echo pin is activated when the sound waves strike the receiver. A computer program can estimate the time gap between the activation of trigger pin and echo pin. This time gap when multiplied by speed of sound which is 343 m/s gives the distance of travel of the sound waves. It may be noted that the distance between the sensor and liquid surface is actually half of the distance calculated using above calculations, as depicted in Fig. 13. Figure 14 shows a simplified algorithm that can estimate the level of liquid in a container and can activate an alarm for low levels of liquid. In this algorithm, as discussed previously a sound wave is generated by triggering transmitter of 10 µs. This time of propagation of sound is calculated in microseconds and stored in the variable “Duration”. Now distance is calculated by multiplying speed of sound in cm/µs with the variable duration and dividing by 2, if this distance is greater than the empty level of liquid set by the user, then low level of liquid in the container will be recorded in the inventory log file and an alarm will activate for 30 s to notify the relevant personal about this shortage. It may be noted that instead of activating an alarm, a valve connected to the container can be opened for some fixed interval of time which will then fill the container with liquid, or a purchasing order can be placed via an automated email generated by the computer program regarding the shortage of goods. Such smart inventory system can be modified depending on the requirement of the industry and ease of the user. Figure 15 displays the basic wiring diagram between an ultrasonic sensor module, buzzer and a microcontroller (Arduino Nano, Atmega 328P). It may be noted that the trigger pin and echo pin are connected to the digital pins 9 and 10 of the microcontroller. The pin 9 connected to the trigger pin is configured as output, while the pin 10 is configured as input pin. For the buzzer, the red wire which is the positive

16

K. Singh

Fig. 13 A visual representation of the estimation of distance using soundwaves

wire of speaker is connected to the digital pin 3 of the controller, with negative pin connected to the ground.

3 Recent Advancements in Inventory Systems Technology has advanced to level where a 3d visualization of goods is possible using laser scanning technology. With appreciable increase in computational processing power in the recent years, technology for such 3d visualizations have not only become cheap, but also compact in size and can be operated with less complications using AI (artificial intelligent) interfaces. The GUI (graphical user interface) for such devices can be accessed using software applications on mobile phones or laptops through cloud based services. There is no doubt that future inventory systems are very difficult to imagine without cloud service based applications for mobile phones or laptops. These systems are unique as they are non-contact type and measure the volume of material inside a container or storage unit by generating a 3d model of the system. This model updates in real time and prevents any financial loss or nonproductive time due to shortage of stock. In this way, the inventory is optimized as the system is able to determine any minute level of changes in the inventory, and goods can be

Smart Inventory Management Using Electronic Sensor …

17

Fig. 14 Algorithm for ultrasonic sensor based smart inventory system

purchased on time. Figure 16 displays such smart inventory system with the capability of visualizing inventory using a 3d model and sending real time information to cloud service which can be accessed through any mobile device or laptop as required. It may be noted that these 3d visualization are best suited for bulk solid inventory, and also for industries where a distance is to be maintained from the stock to maintain hygiene and prevent any damage to product or the sensor. Example, food and pharmaceutical industries pay special attention towards hygiene of their stock. As these sensors are non-contact type they do not interfere physically with the stock material, which increases their operating life and reduces their maintenance cost respectively. Generally, such systems are useful where a pile of stock is created inside the storage bin or the distribution of stock is uneven inside the container as shown in Fig. 16. In such cases it is very difficult to estimate the overall volume of the stock left with the

18

K. Singh

Fig. 15 Basic design of an ultrasonic sensor based smart inventory system

help of other inventory systems like ultrasonic sensor system which is one dimensional in nature and can only estimate the level from a single point on the surface. If we use ultrasonic sensor, the level readings will vary dramatically from point to point on the uneven surface. Load cell based inventory system overcomes this limitation as it calculates the overall weight of the stock and is independent of the level of the surface, however such system is very difficult to install and require proper maintenance with time. As this system has to bear the entire weight of the stock, the structure should be strong which increases the cost of the system. Therefore, a 3d visualization technique is best suited which can measure accurately the volume of stock with the help of complex algorithms, as it can be installed easily and require low maintenance cost as discussed above. Industries that are readily adopting these inventory systems include food processing industries, Cement industries, chemical processing industries, bioenergy industries, coal mining and petrochemical industries, plastic manufacturing industries etc.

Smart Inventory Management Using Electronic Sensor …

19

Fig. 16 Courtesy of DNR Process Solutions Pte Ltd. Image downloaded from https://www.dnrps. com/products-solutions/3d-scanner-multivision/ on 1st June, 2018

From Fig. 17, the action of 3d scanner can be well illustrated. Acoustic pulses are generated from this scanner which are directed at different angles like 15°, 30° and 70° towards the surface of stock as shown. These pulses are reflected back from the surface and are detected by the scanner at different time periods depending upon the topology of the surface. This difference in time period is utilized to generate a three-dimensional model of the surface by sending this data to an HMI or PLC setup. In this way, the volume, level and mass of the stock inside the container can be accurately calculated.

20

K. Singh

Fig. 17 Courtesy of BinMaster design and Manufacturers, a subsidiary company of Garner Industries. Images downloaded from https:// www.azosensors.com/article. aspx?ArticleID=707 and https://www.azosensors.com/ news.aspx?newsID=12271 on 2nd June, 2018

4 Other Possible Sensor Based Inventory Systems As discussed above the idea of combining sensor technology with computer intelligence have a remarkable effect on reducing the cost and improving the efficiency of an industry. However, the sensing technology is not limited to just RFID cards, Ultrasonic sensors, Load cells, or 3d scanners. In fact, some other elegant and inexpensive methods can be employed for development of suitable smart inventory systems. In this section, various such designs as proposed by the author are discussed in detail.

4.1 LDR (Light Dependent Resistor) Based Smart Inventory System LDR act as a photo resistor [10], which is made of a semiconductor with high resistance. The resistance of an LDR or Light detecting resister decreases with increase in the intensity of light falling on its surface. LDR works on the principle of photo conductivity, when light falls on the surface of LDR, the electrons in the valence band of this semiconductor are excited to the conduction band, therefore they are able to conduct electricity when exposed to light. This principle can be utilized to develop a smart inventory system in conjugation with a computer program to analyze the output. Figure 18 displays a system in which a simple crate is designed which has limited number of sections to fit a product in each section. In this example, a crate is designed for 20 products. On the surface of the crate an LDR is placed in each section facing upwards such that it gets covered if a product is placed in a section. In this way placing and removing a product in the crate will affect the resistance of LDR. Covering the LDR with product will increase its resistance to maximum. If the product is removed, then the LDR will be exposed to the ambient light and

Smart Inventory Management Using Electronic Sensor …

21

Fig. 18 LDR (Light Dependent Resistor) based smart inventory system

resistance will decrease immediately. This analog change in resistance value of LDR can be converted to digital output using analog to digital converter and for an array of LDR’s such as in this example it can be converted to a string of 0’s and 1’s using a parallel to serial converter. Using this string of 0’s and 1’s a computer program can not only determine how many products are there in the crate in real time but can also determine which number of section is empty and at what time it was emptied. Such a system in spite of being simple in design is very efficient in determining real time flow of inventory. It will be appropriate to discuss some disadvantages of this system. Even though the system is robust in design and require little maintenance cost, it cannot work if the crate is placed in a dark room. As the LDR’s will not be able to affect their resistances in this condition. Also, this system cannot differentiate between different products, their sizes or weights, as the LDR in a block can be activated by any object that cover its surface. However, such system is very appropriate for a system of inventory where a proper order of similar goods is to be maintained with good lighting conditions.

4.2 IR (Infrared) Transmitter and Receiver Based Smart Inventory System One disadvantage of LDR is that they are difficult to work with in low lighting conditions, to overcome this disadvantage an IR sensor [10] based inventory system

22

K. Singh

Fig. 19 IR (Infrared) Sensor based smart inventory system

can be developed. This sensor has two components one is IR transmitter and other is IR receiver. Transmitter is an Led which transmits infrared light and receiver is a photodiode which is sensitive to the infrared light. Usually both the transmitter and receiver are paired together and face in same direction. Infrared light from the transmitter strike the surface of an object nearby and the reflected light is sensed by the receiver, in this way an object in front of this sensor module can be detected. Figure 19 displays an inventory system designed on the similar principle of object detection using IR sensors. A crate consists of a horizontal array of IR sensors as shown. When an object is placed in one of the section of the crate, IR sensor of that particular section gets activated and sends a high output signal. Again, the output of all the IR sensor modules can be converted from parallel to serial data and can be analyzed by a computer program for real time monitoring of the inventory system. In this system, we do not require analog to digital converter as an IR sensor module have an inbuilt operational amplifier which outputs a digital high when IR receiver is activated.

5 Wireless Technologies to Improve Portability of Smart Inventory Systems Portability plays a vital role in a smart inventory system. A portable system has an advantage of being compact and user friendly. Also, it is much easier to install. As discussed above, web servers and cloud based services can be utilized for accessing real time information globally, which is one way of accessing the inventory system. Other wireless technologies such as Bluetooth [11], or radio communication can be utilized for accessing information locally. Such technologies have the advantage of being more reliable than cloud services as the information is being transferred directly from the inventory system to the GUI interface instead of going through the web server. The disadvantage of such technologies is their short range of commu-

Smart Inventory Management Using Electronic Sensor …

23

Fig. 20 Three different wireless solutions for smart inventory systems

nication. Example, Bluetooth have a range of approximately 100 m. One example of a Bluetooth is HC-05 module which can be configured as master or slave i.e., it can be configured to transfer data or receive data. This module can communicate with a controller or a computer using UART [12] (Universal Asynchronous receiver/transmitter) communication protocol. It doesn’t require a clock signal to communicate data with the controller. Data is communicated via the transmitter and receiver pins of the module. If we consider radio communication for our inventory system, its range depends on the system used, Example 433 MHz [13] radio transmitter and receiver modules mainly used for digital communication can deliver a range of 100–200 m safely without any interference depending upon the antenna used. Other example of radio communication module that can be used is NRF24L01 module, which operates at a frequency of 2.4 GHz. This module has the advantage that it can act both as transmitter or receiver, therefore there is communication possible in both the directions, also due to high frequency range of 2.4 GHz, more information can be transferred at much higher rate and bandwidth. This module uses SPI [12] (Serial Peripheral interface) communication protocol to communicate with a controller or a processor. This protocol is different than UART as it requires a common clock frequency to communicate data. Figure 20 displays the wireless modules as discussed above which are inexpensive but effective for wireless communication if utilized in conjugation with smart inventory systems as discussed in this chapter.

6 Conclusion In this chapter, we discussed how different smart inventory systems which are a combination of electronic sensor technologies and computational intelligence can effectively enhance the productivity of an industry or business in terms of customer

24

K. Singh

satisfaction and effortless management of inventory. Using such systems, the overall efficiency of an industry increases, as the requirement of stock can be predicted using valuable data obtained from these inventory systems. Also, as discussed above there are various technologies that can be adopted for a smart inventory system. Every technology has its own advantages and drawbacks. The selection of an inventory system depends on the type of product and work environment of the industry. From the above discussion, it can be concluded that RFID technology is best suited for packaged products, and can improve the overall management of the stock, if adopted with proper infrastructure. It may also be concluded that Load cell technology can be adopted both for packaged or non-packaged products using appropriate algorithm as discussed in this chapter. The disadvantage of load cells being the physical stresses generated within load cell, which require them to be structurally strong to bear the weight of the stock. Ultrasonic technology has been discussed thoroughly, from which we may conclude that its best suited for liquid stock stored in containers and is preferred due to its non-contact nature and low maintenance cost, however it is not suitable for stock such as powders, grains, chemicals etc. which bears uneven topology in the storage container. For such systems 3d scanners are preferred which map the surface of the stock by sensing the sonic waves propagated at different angles towards the stock surface. It may be observed that some of the inexpensive solutions like use of LRD’s and IR sensor modules may prove to be as efficient as other sensor based technologies if properly designed and implemented as discussed above. From the discussion of wireless technologies such as Bluetooth and radio communication it may be concluded that such technologies can be adopted widely to improve that portability of a smart inventory system. As discussed above these wireless solutions are better than server based systems if adopted for local communication as they send the data directly for the inventory system to the GUI interface without any intervening server as a medium.

References 1. Ramaekers, K., Janssens, G.K.: Modelling the complexity of inventory management systems for intermittent demand using a simulation-optimisation approach. In: Aziz-Alaoui, M.A., Bertelle, C. (eds.) From System Complexity to Emergent Properties. Understanding Complex Systems. Springer, Berlin, Heidelberg, pp. 303–313 (2009). Print ISBN 978-3-642-02198-5 2. Zjajo, A., Gyvez, J.P.D.: Low-power high-resolution analog to digital converters. In: From Analog Circuits and Signal Processing. Springer, Netherlands, pp. 11–40 (2011). eBook ISBN 978-90-481-9725-5 3. Qiao, Y., Chen, S.S., Li, T.: RFID as an Infrastructure. From SpringerBriefs in Computer Science, pp. 1–8. Springer, New York (2013). eBook ISBN 978-1-4614-5230-0 4. Zangl, H., Bretterklieber, T.: Elektrotech. Inftech. 124, 364 (2007). https://doi.org/10.1007/ s00502-007-0481-0 5. Seemann, K., Huemer, M.: Elektrotech. Inftech. 122, 452 (2005). https://doi.org/10.1007/ BF03054377 6. Han, J., Xi, W., Zhao, K., Jiang, Z.: Device-Free Object Tracking Using Passive Tags. From SpringerBriefs in Electrical and Computer Engineering, pp. 2–3 (2014). eBook ISBN 978-3319-12646-3

Smart Inventory Management Using Electronic Sensor …

25

7. Mihulowicz, P.J., Weglarski, M., Pitera, G., Kawalec, D., Lichon, W.: Development board of the autonomous semi-passive RFID transponder. Bull. Pol. Acad. Sci. Tech. Sci. 64(3) (2016). https://doi.org/10.1515/bpasts-2016-0073 8. Gupta, S.V.: Strain gauge load cells. In: Mass Metrology. Springer Series in Materials Science, vol. 155. Springer, Berlin, Heidelberg (2012) 9. Ihara, I.: Ultrasonic sensing: fundamentals and its applications to nondestructive evaluation. In: Mukhopadhyay, S., Huang, R. (eds.) Sensors. Lecture Notes Electrical Engineering, vol. 21. Springer, Berlin, Heidelberg (2008) 10. McGrath, M.J., Scanaill, C.N.: Sensing and sensor fundamentals. In: Sensor Technologies. Apress, Berkeley, CA (2013) 11. Vladimerou, V., Dullerud, G.: Wireless control with bluetooth. In: Hristu-Varsakelis, D., Levine, W.S. (eds.) Handbook of Networked and Embedded Control Systems. Control Engineering. Birkhäuser, Boston (2005) 12. Subero, A.: USART, SPI, and I2C: serial communication protocols. In: Programming PIC Microcontrollers with XC8. Apress, Berkeley, CA (2018) 13. Yang, Q., Huang, L.: 433/315 MHz communication. In: Inside Radio: An Attack and Defense Guide. Springer, Singapore (2018)

Computational Intelligence for Data Analytics S. Anupama Kumar, M. N. Vijayalakshmi, T. L. Divya and K. N. Subramanya

Abstract Computational Intelligence is one of the state of art technology which can be widely used in various applications. Data analytics is one prime area which is implemented in various domains to help the society grow better. The outcome of the analytics helps the decision makers to make better decisions and improve the business. This chapter brings in the implementation of computational intelligence through different machine learning algorithms. Topic modelling is implemented over customer review data set to generate terms and topics to analyse the review and understand the behaviour of the customers towards the product. Various classification algorithms were applied over an educational dataset to analyse the performance of the students and help the tutor to make decisions in changing the course structure. An experimental setup was made to make the algorithms learn the dataset through previous records and then new records were introduced. The model is then evaluated using different metrics and the best model is identified for the selected dataset. This chapter is an application of the above mentioned techniques to perform data analytics in a better way. Keywords Machine learning · Educational data mining · Topic modelling · LDA · Beta metrics · Preplexity efficiency · ROC · PRC · Rank

1 Introduction Computational Intelligence (CI) is study of intelligent agents which reacts appropriately under various circumstances. The intelligent system is adaptable to change in environments and learns from experiences. It can make appropriate choices even in precincts and finite computation. The principal scientific goal of computational intelligence is to understand the principles that make intelligent behavior possible, in natural or artificial systems [1]. The science of CI is called as “computational S. Anupama Kumar (B) · M. N. Vijayalakshmi · T. L. Divya · K. N. Subramanya Department of MCA, R V College of Engineering, Bengaluru, India e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_2

27

28

S. Anupama Kumar et al.

epistemology”—the study of knowledge. It can be seen as a way to study the existing problem and provide solution using a powerful experimental tool. CI allows us to experiment with executable models of intelligent behavior and bring the hidden knowledge from it. These models are open to inspection, redesign, and experiment in a complete and rigorous way. Computational intelligent techniques can be widely used to solve and overcome the challenges that arise in data analysis. Data analysis becomes a complicated technique due to the volume of data, the heterogeneity of the data, nature of the inter dependencies among the data and unknown disturbances. Computational intelligence are designed to meet these challenges with the help of machine learning algorithms and tools which can handle huge amount of data, learn from previous experience and build a model using appropriate techniques. It can also analyze the effectiveness and efficiency of the model using appropriate metrics. The applications of computational Intelligence are diverse, viz medical diagnosis, scheduling resources and processes, designing robots for hazardous environments, games, autonomous vehicles, natural language translation systems etc. Computational systems can be designed using machine learning algorithms and data analytical tools as an automated system which is capable of handling huge amount of heterogeneous data, learn from previous experience and deliver an accurate model using the learned experience. This chapter discusses two data analytical applications (i) Application of topic modeling to understand the behavior of customers to improve business and (ii) Analyzing the performance of students in a higher educational system using machine learning techniques. An experimental model is designed and built for both the applications which are capable of handling the varied data from different sources. Different machine learning algorithms were applied over the datasets and the efficiency of the algorithms were measured using appropriate metrics. Topic modeling is a text mining technique of discovering “topics” from a collection of documents. It can be applied to discover the method in which a document is structured and bring out the unknown information from them. It helps us to organize and provide deep insights into the large unstructured collection of text bodies and infer new knowledge from it. The applications of topic modeling ranges from emails, social data including twitter and Facebook, product reviews, health care analysis, bio informatics etc. LDA algorithm has been implemented using R tool to identify the topics and analyze the customer reviews. Two different metrics has been implemented to evaluate the model that is best suited for the analyzing the topics. Educational Data Mining (EDM) models were built to analyze the course outcomes of the students in a higher education system. The output of the model would help the course coordinators to understand the behavior of the student under various learning methods and the learning outcome of the learner in the course. This would also help them to make revisions in the course structure and bring out betterment in the education system. Various classification algorithms like random forest, rep tree, Random tree and J48 has been implemented using WEKA tool to analyze the student performance and the best model has been identified using various metrics.

Computational Intelligence for Data Analytics

29

The following section gives an insight into Topic Modeling, data preprocessing, implementation of the algorithm, the results of the model and the efficiency of the algorithms in detail.

2 Topic Modeling for Customer Review Analysis The analysis of the text content in social media like emails, blogs, tweets, forums etc. of textual communication constitutes text analytics. Text analytics helps to analyze millions of data tweeted by customers’ comments and questions in forums; and performs behavioral analysis using different algorithms. Topic modeling can be described as a method for finding a group of words (i.e. topic) from a collection of reviews/documents that represents the key information from that collection. This can also help us to organize, structure and obtain a recurring pattern of words from a text document. The main objectives of topic modeling are: • Discovering hidden topical patterns that are present across the collection • Annotating documents according to these topics and use these annotations to organize search and summarize texts. Topic modelling can be applied in the following areas of research. Text classification—To improve classification of the documents by grouping similar words together in topics rather than using each word as a feature Recommender Systems—Recommender systems can be built using similarity measures Uncovering Themes in Texts—Modelling methods can be used to detect trends in market for a particular product. There are several algorithms for doing topic modelling. The most popular ones include • LDA—Latent Dirichlet Allocation—They are Probabilistic Graphical Models • LSA or LSI—Latent Semantic Analysis or Latent Semantic Indexing—Uses Singular Value Decomposition (SVD) on the Document-Term Matrix. Based on Linear Algebra • NMF—Non-Negative Matrix Factorization—Based on Linear Algebra [2].

2.1 Experimental Setup This section gives an insight into the implementation of topic modelling to analyse the customer review dataset on apple i-phones. The analysis will help the manufacturers to understand the expectation of the customer towards the product and improvise the necessary features. Topic models are built using the following assumptions • The reviews consists of a mixture of topics and • Each topic consists of a collection of words.

30

S. Anupama Kumar et al.

Fig. 1 Experimental set up for topic modelling

The models are built under the assumption that the semantics of the reviews are governed by hidden, or “latent,” variables that are not observed. The uncovered variables bring topics—that shape the meaning of corpus. The customer review dataset consists of unstructured text data pertaining to the reviews of the customers towards apple i-phones. There are a total of 155 reviews available in the data set. The unstructured data has to be pre-processed so that it can be analysed further. The Fig. 1 shows the experimental setup of the model designed to analyse the customer reviews. The initial step in analysing the customer reviews is to pre-process the unstructured data so that the text is ready for implementation. A sample review is shown below from the original review dataset

It can be observed from the review that it contains a lot of unwanted text which has to be pre-processed to make be ready for analysis. The pre-processing steps are explained as below 1. Convert the document into a corpus 2. Convert the document into lowercase so that the document become easy to do further processing 3. Remove the stops word which does not make sense in the sentence. The system should be instructed to remove the stop words which are present in English with respect to this dataset 4. Remove the punctuation symbols like ,:. Etc and the numbers if they are present in the review

Computational Intelligence for Data Analytics

31

5. Stemming is the process of removing the stems from the words. The following statement is an example of the pre processed document after removing the punctuation and numbers. From the output generated it’s understood that the punctuation marks and numbers are removed. In the same way white spaces will also be stripped out.

The statement below is the output of the reviews after the pre-processing. The whole review cannot be viewed due to its size, but the system displays the meta data of the document.

As a final step in pre-processing a word cloud has been formed out of the most frequently occurring data. The following figure shows the word cloud of the top 10 words that has occurred in the review data set after pre processing (Fig. 2). From the word cloud it is easily understood that ipod (present in black color) is the most frequently used word, followed by nano, will, video, can and must (present in pink color). The next common words were good, apple, sound, quality etc. and then player, new, camera, play etc. This word cloud helps the user to understand what are the common topics that might occur in the review set. The following section explain the implementation of the LDA algorithm

Fig. 2 Word cloud of the terms generated

32

S. Anupama Kumar et al.

2.2 Implementation of LDA Algorithm to Analyse Customer Reviews Latent Dirichlet Allocation (LDA) is a generative and probabilistic model that can be used to automatically group words into topics and documents into a mixture of topics [3]. In [4] the author has given an insight into the working of the different topic modelling algorithms like LSA, LDA, PLSA etc. The following Fig. 3 discusses the working pattern of LDA algorithm [5]. Each word comes from different topics present in the review. Therefore it becomes necessary to form a document term matrix to understand the structure of the terms. The term document matrix lists out the number of documents present (in this case the reviews are stored as a single document), the number of unique terms, the sparsity, and maximal term length. The following data is the output of the Document Term matrix generated for the review dataset.

The sample terms with the total number of frequencies is displayed. The following terms are found to be frequently appearing in the review dataset.

Fig. 3 Implementation steps for LDA algorithm

Computational Intelligence for Data Analytics

33

The sum of words is found for the document term matrix for further implementation.

The LDA algorithm is implemented for 2 topics and 4 topics using the Variational Expectation maximization (VEM) methods. The following is the output of the implementation of the algorithm using the beta metrics.

To understand the output of the LDA algorithm, a graph has been plotted with the topics generated. The top three terms generated under each topic is common and the other terms are different in the topics (Fig. 4). The algorithm is again iterated by increasing the number of topics to 4 without changing the other parameters. The topics listed out from the algorithm is as follows

34

S. Anupama Kumar et al.

Fig. 4 Result generated for 2 topics with top 10 terms

The number of terms under each topic is increased by the implementation of the algorithm. The output of the algorithm is depicted as a graph as shown in Fig. 5. From both the graphs generated using LDA algorithm, it is clear that the terms product and recommend has got higher priority when the number of topics was 2 and 4. Therefore it is concluded that the review towards the product is positive.

Computational Intelligence for Data Analytics

35

Fig. 5 Result generated for 4 topics with top 10 terms

2.3 Analysis of LDA Algorithm Over Customer Review Data Set The next step is to turn the model into a one-topic-per-term-per-row format. For each combination, the model computes the probability of the term being generated from that topic. To evaluate the accuracy of the algorithm used beta spread method can be applied to find the greatest different between topic 1 and topic 2. The difference can be estimated based on the log ratio of the two topics generated. The following table shows the differences between the terms which are commonly present in both the topic.

36

S. Anupama Kumar et al.

To constrain the relevance to frequent words, we can filter for relatively common words, such as those that have a $\beta$ greater than 1/1000 in at least one topic. The words with greatest difference are better and least difference is bad. The following graph is the visualization of the beta spread of the reviews. The application of log2 analysis for 2 topic relevance is shown in Fig. 6. The visualization states clearly the log2 differences in the terms that are present in the topic 2 models. The efficiency of the algorithm towards the given dataset can be handled using the statistical measure called Perplexity [6]. For a given value of K, the LDA model can be estimated by taking the word distributions represented by the topics and compared with the actual topic mixtures of the document. The benefit of this statistic comes in comparing perplexity across different models with varying k’s. The model with the lowest perplexity is generally considered the “best”. The perplexity values of k  2 and k  4 are compared to evaluate the model. The perplexity values of the model are given below

The perplexity value of the LDA model when k  2 is less than the value when k  4. There the model with k  2 is better suitable for the given dataset. The experiment is found successful in applying LDA algorithm for k  2 in generating two unknown topics and the various terms related to it. The analysis of the term infers a positive review towards the product.

Computational Intelligence for Data Analytics

37

Fig. 6 Analysis of the result using Log2 ratio for 2 topics

3 Machine Learning for Educational Data Analysis Educational research is one leading research areas worldwide. Educational brings in new technological advancements to improvise the society. A large number of research methods are available to analyze educational data pertaining to student performance, interest towards various learning environments, teaching –learning methods, improving the course contents, structure etc. In [7] Educational process mining (EPM) is defined as an emerging field in educational data mining (EDM) aiming to make unexpressed knowledge explicit and to facilitate better understanding of the educational process. The educational process is facilitated by different stakeholders and a lot of challenges arise in the process. The authors in [8] have stated the different research works that has emerged over the years in the field of education data mining and the applications of various techniques and methods in that. In this chapter an Educational process model has been built to collect the data, process and analyze it to bring out betterment in education.

38

S. Anupama Kumar et al.

3.1 Experimental Setup This section gives an understanding towards the experimental setup which collects the student performance data, implements different machine learning algorithm, analyzes the algorithms for its efficiency and effectiveness and finds the model which best suits for the dataset. The main objective of this model to predict and analyze the course outcome of the students based on their performance under various learning methods using machine learning algorithms. Further the efficiency and effectiveness of the algorithms are also analyzed to find the best suited model for this dataset. Figure 7 is the experimental set for the computational model built. The dataset used in this experiment comprises of the student performance data including the marks scored by the student under various learning methods like multiple quiz, assignment, tests and practicals. The attendance percentage of the student in the class is also considered as a factor to analyze the performance of the student. The test and practical marks includes the marks scored by the student in each question/labcycle and then consolidated for total marks. Preprocessing the dataset is one of the important phases in this experiment. This enabled to handle the missing values and noise that was present in the dataset. The missing values in this case are the marks of the questions in quiz/test which the student has not attended and therefore are replaced by zero. The marks of the student who has not attended the tests has been marked as AB and is replaced by zero for the purpose of working. The target value of the dataset is YES/NO stating whether the student achieves the course outcome levels or not. In case the student fails to meet it, modifications have to made in the course structure. Therefore it is necessary to find the percentage of students who achieves the target value by analyzing their performance records. The following section gives a detailed view of the implementation of the classification techniques used in this experiment.

3.2 Implementation of Classification Techniques Cristóbal Romero [9] in his article has explained the importance of Educational data mining and the various tasks associated with it. He has also explained how classifica-

Fig. 7 Experimental set up of educational data model

Computational Intelligence for Data Analytics Table 1 Implementation of classification techniques

39

Sl no

Name of the algorithm

No of records correctly predicted

No of records incorrectly predicted

1

J48

100

13

2

Random forest

99

14

3

Random tree

100

13

4

RepTree

101

12

tion techniques can be applied over the educational datasets to infer new knowledge from it. In [10] the authors have explored different classification techniques and found the techniques to be successful in predicting the student performance. In this research work, various classification techniques based on decision trees have been implemented to analyze the course outcome of the students in a post graduate higher education system. The authors in [11] have used different decision tree classification techniques to predict the performance of the students and concluded REP tree to be one of the best suited algorithm. Four classification trees like J48, Random tree, REP tree and Random forest has been implemented to analyze the data set. A 10 fold cross validation is used to train and test the data set for better results. The following output has been generated from all the trees. The Table 1 lists the numberζ of correctly and incorrectly predicted records for the target values. From the Table 1 it’s understood that REP tree has predicted more records correctly than the other records. The following Fig. 8 shows the depiction of J48 and Fig. 9 shows the depiction of Random tree generated after the implementation of the algorithm over the dataset.

Fig. 8 Implementation of J48 algorithm

40

S. Anupama Kumar et al.

Fig. 9 Implementation of random tree algorithm

The attributes that contribute to predict the course outcome are the total marks obtained by the students in each test, their course outcomes at each level. At each level of the tree it clearly states that the analysis of the attribute leads to the target value (YES/NO). The total number of records which are correctly predicted using J48 is 100 and 13 records are incorrectly predicted. The random tree algorithm predicted 100 records correctly and 13 records incorrectly like the J48 algorithm. But when the output trees are analyzed, the attributes taken by both the algorithms in their root and leaf nodes differ from each other. All these attributes are necessary to reach the target. From the Table 1, we can conclude that REP tree will be the most suitable algorithm for this model. But the percentage of accuracy may not be only parameter to select the algorithm. The following section discusses the various metrics are used to analyze the efficiency of the algorithms used and decide upon the accurate model.

3.3 Evaluation of Classification Techniques Evaluating the classification algorithm to analyze the student performance records is a vital step in this experiment. The author in [12] has explained the different techniques that can be used. The author in [13] have discussed various metrics that can be used to analyze the classification algorithms. The application of these techniques over the educational data set is also discussed. In this experiment the classification accuracy

Computational Intelligence for Data Analytics

41

is estimated using the Accuracy rate, error rate, ROC, F-measure and time taken to build the model. The classification measures implemented above are defined as below 1. Accuracy: It describes the number of correct predictions against the total number of predictions. Accuracy  (f11 + f00 )/(f11 + f00 + f10 + f01 )

(1)

2. Error rate: It explains the number of wrong predictions against the total number of predictions Error rate  (f10 + f01 )/(f11 + f00 + f10 + f01 )

(2)

3. F Measure: F Measure  (2(precision ∗ recall))/(precision + recall) 4. ROC Value: The ROC curve is a metric that is plotted from the origin which moves for each true positive and false positive values [14]. The vertical and horizontal axis shows the percentage of true positives and false positives. Any ROC curve value holding between 0 and 1 is highly appreciative for a better accuracy. The following Table 2 shows the classification accuracy of all the four algorithms over the metrics. From the Table 2 it is clear that J48 and Random tree performs in the similar way of predicting even though they have used a different attribute set to frame the tree. Random forest is not suitable since the accuracy rate is low even after taking more time to build the model. The REP is the more suitable tree to be used in this model since it has a higher accuracy rate, low error rate. The other factors like F measure and ROC are within the limits. The more time taken by the model against the J48 and Random tree can be justified since the accuracy rate of the model is high. Finally it is evident that the classification models can perform well to evaluate the student performance records and identify the achievement of the course outcome.

Table 2 Evaluation of classification techniques Sl no

Name of the algorithm

Accuracy rate

Error rate

ROC

F measure

Time taken (s)

1

J48

88.49

0.11

0.683

0.882

0

2

Random forest

0.1232

0.918

0.843

0.34

3

Random tree

88.49

0.11

0.683

0.882

0

4

RepTree

89.3805

0.1062

0.865

0.531

0.02

0.882

42

S. Anupama Kumar et al.

For the given dataset since the number of student who achieve the course outcome (target value  Yes) is more and predicted correctly by the algorithm, the tutor is advised not the make major modifications or re structure the course outlines.

4 Conclusion Computational intelligence techniques were combined with the machine learning algorithms to perform data analysis for two different applications. The experimental setup was tested under different conditions and evaluated for various values. The implementation of the LDA algorithm has generated various topics and was tested using prevalence metrics to understand the efficiency of the model generated. From the terms generated under the topics it is understood that the product has a positive review from the customers. The educational data set has been analysed using the classification algorithms and found to be effective in predicting the course outcome of a post graduate course using various attributes. The algorithms were evaluated using its accuracy rates and time taken to build the model. From the output generated it is understood that the students were able to attain the course outcome and REP tree was the most suited classification technique to run the model. In future computational techniques can be widely used in various applications like health care sector, image and video analysis, financial sector etc. to bring in betterment to the society.

References 1. Poole, D., Mackworth, A., Goebel, R.: Computational Intelligence A Logical Approach. Copyright © Oxford University Press, New York (1998) 2. Complete Guide to Topic Modelling. https://nlpforhackers.io/topic-modeling/ 3. Blei, D.M., Ng, A.Y., Jordan, M.I.: Latent dirichlet allocation. J. Mach. Lear. Res. 3, 993–1022 (2003) 4. Xu, J.: How to easily do topic modeling with LSA, PLSA, LDA & lda2Vec. Stanford Universtiy. https://medium.com/nanonets/topic-modeling-with-lsa-psla-lda-and-lda2vec 5. Miran, S.: Latent Dirichlet Allocation (LDA) for Topic Modelling. Project presentation. http:// www.ece.umd.edu/smiran/LDA.pdf 6. Moreno, A., Redondo, T.: Text analytics: the convergence of big data and artificial intelligence. Int. J. Interact. Multimedia Artif. Intell. 3(6), 57–64 (2016) 7. Bogarín, A., Cerezo, R., Romero, C.: A survey on educational process mining. First published 28 Sept 2017. https://doi.org/10.1002/widm.1230 8. Sin, K., Muthu, L.: Application of big data in education data mining and learning analytics—a literature review. ICTACT J. Soft Comput. Models Big Data 5(4), 1035–1049 (2015) 9. Romero, C., Ventura, S.: Educational data mining: a review of the state of the art. IEEE Trans. syst. Man Cybern. Part C Appl. Rev. 40(6) (2010) 10. Danso, S.O.: An exploration of classification prediction techniques in data mining: the insurance domain. A Dissertation Presented to the School of Design, Engineering, and Computing. Bournemouth University. http://www.comp.leeds.ac.uk

Computational Intelligence for Data Analytics

43

11. Anupama Kumar, S., Vijayalakshmi, M.N.: Efficiency of decision trees in predicting student’s academic performance. In: Proceedings of First International conference on Computer science, Engineering and Applications (CCSEA 2011), Chennai, pp. 335–341, 15–17th July 2011, ISSN: 2231-5403 12. Ukwueze Frederick, N., Okezie Christiana, C.: Evaluation of data mining classification algorithms for predicting students performance in technical trades. Int. J. Eng. Comput. Sci. 5(8), 17593–17601 2016. ISSN: 2319-7242 13. Anupama Kumar, S., Vijayalakshmi, M.N.: Efficiency of multi instance learning in educational data mining. In: Margret Anouncia, S., Wiil, U. (Eds.) Knowledge Computing and its Applications—Volume II. Springer, Singapore (2018). Print ISBN: 978–981-10-8257-3, Online ISBN: 978-981-10-8258-0 14. Ramesh, V., Thenmozhi, P., Ramar, K.: Study of influencing factors of academic performance of students: a data mining approach. Int. J. Sci. Eng. Res. 3(7) (2012) 15. Kireyev, K., Palen, L., Anderson, K.: Applications of topics models to analysis of disasterrelated twitter data. In: NIPS Workshop on Applications for Topic Models: Text and Beyond, vol. 1 (2009) 16. Hospedales, T., Gong, S., Xiang, T.: A Markov clustering topic model for mining behaviour in video. In: 2009 IEEE 12th International Conference on Computer Vision. IEEE (2009) 17. McAuley, J., Leskovec, J.: Hidden factors and hidden topics: understanding rating dimensions with review text. In: Proceedings of the 7th ACM Conference on Recommender Systems. ACM (2013) 18. Rehurek, R., Sojka, P.: Software framework for topic modelling with large corpora (2010)

Design and Analysis of an Enhanced Multifactor Authentication Through a Covert Approach Raman Kumar and Uffe Kock Wiil

Abstract Most of the network service providers currently use two-factor authentication for their wireless networks. This exposes the network subscribers to identify theft and user’s data to security threats like snooping, sniffing, spoofing and phishing. There is need to control these problems with the use of an enhanced multi factor authentication approach. The objective of this work is to create a multi-factor authentication software for a wireless network. Multi factor authentication involves user’s knowledge factor, user’s possession factor and user’s inherence factor; that is who the user is to be presented before system access can be granted. Multi factor authentication depends mainly on three factors: (1) Something a user knows, such as a password or PIN (2) Something a user has, such as a key, a card, or another kind of token (3) Something a user is, such as a retina scan, or fingerprint. We may enhance the reliability and security of the authentication mechanism by combining multiple authentication factors into a single model. Multi factor authentication is far better if we use this schema both statically and dynamically. The three factors together provide a much higher confidence in the all prerequisite parameters of cryptography. Keywords Multifactor authentication · One factor · Two factor · Authentication and authorization

R. Kumar (B) Department of Computer Science and Engineering, I K Gujral Punjab Technical University, Kapurthala, Punjab, India e-mail: [email protected]; [email protected] U. K. Wiil The Maersk Mc-Kinney Moller Institute, University of Southern Denmark, Odense, Denmark e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_3

45

46

R. Kumar and U. K. Wiil

1 Introduction In this paper, various parameters related to multifactor authentication (MFA) are discussed. Previous authors have discussed about schema statically. MFA has more factors as compare to authentication like one factor, two factor, etc. The covert approach is an enhanced technique to implement authentication using added dimension of digital signature (see Fig. 1). This approach can work both statically and dynamically. MFA requires factors from two or more of the following categories: • Users knowledge • Something a user has • Biometric that’s integral to the user (e.g., fingerprint, iris, voice, etc.). Figure 2 demonstrates how two-factor authentication works. Figure 3 shows the proposed multifactor authentication trough a covert approach. The paper is organized as follows. Section 1 gives a brief overview of multi factor authentication. In Sect. 2, we provide an overview of related work presenting existing techniques. Section 3 introduces the proposed work with added features

Fig. 1 A diagrammatical view of multifactor authentication via covert approach

Fig. 2 A layout for two-factor authentication

Design and Analysis of an Enhanced Multifactor …

47

Fig. 3 A layout for multifactor authentication via covert approach

such as static and dynamic auditing confirmation. In Sect. 4, we demonstrate the results of an analysis conducted on the proposed solution. Section 5 concludes on the conducted work and its outcomes.

2 Related Work There have been many smart-card-based password authentication schemes suggested in the previous literature [1–7]. In Lee et al.’s scheme [7], two password-based two-factor authentication and key exchange protocols are proposed. The first protocol does not provide pseudo identity and the second protocol provides identity protection. Both protocols require only two messages exchanging. These proposed protocols are suitable for low-power devices such as PDAs in public wireless LANs which require mutual authentication, low computation cost, identity protection, and less exchanged messages. In Hwang et al.’s scheme [2], a secure mutual authentication method is introduced. In Wu and Zhu’s scheme [8], a secure authenticated key exchange protocol is presented that achieves fully two-factor authentication and provides forward security

48

R. Kumar and U. K. Wiil

of session keys. They have used user’s unique identity to accomplish authentication, instead of using public keys. They used nonces instead of timestamps to avoid the clock synchronization problem. Their scheme allows users to change their password freely without any interaction with the server. They have also given a security proof of their protocol using random-oracle model. Das [9] proposed a two-factor user authentication protocol for WSN using only hash function. The proposed protocol avoids many logged in users with the same login-id and stolen-verifier attacks, which are prominent threats for a password-based system if it maintains verifier table at the GW node or sensor node. In addition, the proposed protocol resists other attacks in WSN except the denial-of-service and node compromise attacks. They have showed the efficiency of the proposed protocol in comparisons with the related ones. Khan and Alghathbar [10] have shown in their scheme that a recently proposed two-factor user authentication scheme in WSN environment is insecure against different kinds of attack and should not be implemented in real applications. They have demonstrated that in the Das scheme [9], there is no provision for users to change or update their passwords, the GW node bypassing attack is possible, it does not provide mutual authentication between GW node and sensor node, and it is susceptible to privileged-insider attack. To remedy the before mentioned flaws, they have proposed security patches and improvements, which overcome the weak features of the Das scheme. The presented security improvements can easily be incorporated in the Das scheme for a more secure and robust two-factor user authentication in WSNs. Nyang and Lee [11] pointed out that Das’ [9] two-factor user authentication protocol is weak against the off-line password guessing attack by insiders and showed that a simple patch that appends secret parameter to the authentication information can eliminate this weakness without sacrificing any efficiency and usability. Also, to protect query responses from wireless sensor nodes to a user, they proposed an efficient method which can be easily implemented using a built-in AES function in sensor nodes. Finally, they gave a guideline for secure implementation of authentication protocols, which prevents the outsider who captures a sensor node from mounting password guessing attack and from impersonating the gateway node. Pu [12] suggested that in addition to the five desirable properties (client authentication, server authentication, server knows no password, freedom of password change and prevention from guessing attack), key compromise impersonation resilience should also be added as one more important security requirement for two factor smart-card-based password mutual authentication [6]. It means the adversary should not be able to masquerade any user to access the server’s service once if the long-term key of the server is compromised. They provided an attack to illustrate the adversary is able to masquerade any user to access the server’s service in their protocol once if the long-term key of the server is compromised. Finally, they have proposed such an improved protocol that eliminates the security weakness existing in Yang et al.’s protocol [6] i.e. allowing key-compromise impersonation.

Design and Analysis of an Enhanced Multifactor … Table 1 Notions [1]

Symbol

49

Physical meaning

sk t

Secret tag key

pkt

Public tag key

sk h

Secret hash kev

M

Data component

T

Set of data tags

N

Number of blocks in each component

S

Number of sectors in each data block

M info

Abstract information of M

C

Challenge generated by the auditor

P

Proof generated by the server

Fig. 4 System model of the data storage auditing [1]

3 Proposed Work We start with an overview of existing methods and continue with a presentation of the proposed enhancements to the current solutions. Hwang [1] has developed a method for secure and optimal authentication using multifactor authentication. There are many attacks that can breach into security credentials and some of the dangerous attacks are reply, forge attack. Hwang [1] introduces an index table to record the abstract information of the data (Table 1) (Fig. 4). Cloud computing is a promising computing model that enables convenient and on-demand network access to a shared pool of computing resources. Cloud storage is an important service of cloud computing, which allows data owners to move data from their local computing systems to the Cloud. More and more data owners start

50

R. Kumar and U. K. Wiil

choosing to host their data in the Cloud [3]. The local management of such huge amount of data is problematic and costly due to the requirements of high storage capacity and qualified personnel. Therefore, Storage-as-a-Service offered by cloud service providers (CSPs) emerged as a solution to mitigate the burden of large local data storage and reduce the maintenance cost by means of outsourcing data storage [4]. Cloud Storage Providers like Microsoft with Sky Drive, Google Documents, and Drop Box etc. have successfully dropped rates of storage available on internet. They promise availability of the data from different systems/locations/networks. Basic security like User based authentication access of data and maintaining offline data to the client’s machine is also supported [4]. Owner initialization: The owner runs the key generation algorithm KeyGen to generate the secret hash key skh , the pair of secret-public tag key (skt ; pkt ). Then, it runs the tag generation algorithm TagGen to compute the data tags. After all the data tags are generated, the owner sends each data component M  {mi}i [1 ,n ] and its corresponding data tags T  {ti }i[1,n] to the server together with the set of parameters {uj }j[1,s] . The owner then sends the public tag key pkt , the secret hash key skt , and the abstract information of the data Minfo to the auditor, which includes the data identifier FID, the total number of data blocks n. Static Auditing (Fig. 5). Dynamic Auditing (Fig. 6). Data update (Phase 1): There are three types of data update operations that can be used by the owner: modification, insertion, and deletion. For each update operation, there is a corresponding algorithm in the dynamic auditing to process the operation and facilitate the future auditing, defined as follows: Step 1: Modify (mi , skt , skh ) → (Msgmodify , ti* ). The modification algorithm takes as inputs the new version of data block m*i , the secret tag key skt , and the secret hash

Fig. 5 Phase 1 (static owner initialization)

Design and Analysis of an Enhanced Multifactor …

51

Fig. 6 Dynamic data update (Phase 1)

key skh . It generates a new version number V*i , new time stamp Ti* , and calls the TagGen to generate a new data tag ti* for data block m*i . It outputs the new tag t *i and the update message Msgmodify  (i, B*i ,V *i ,T *i )Then, it sends the new pair of data block and tag(m*i , t *i ) to the server and sends the update message Msgmodify to the auditor. Step 2: Insert(m*i , skt , skh ) → (Msgmodify , ti* ). The insertion algorithm takes as inputs the new data block m*i , the secret tag key skt , and the secret hash key skh . It inserts a new data block m*i before the ith position. It generates an original number B*i , a new version number V*i , and a new time stamp T*i . Then, it calls the TagGen to generate a new data tag t*i for the new data block m*i . It outputs the new tag t*i and the update message Msginsert  (i, B*i ,V *i ,T *i ). Then, it inserts the new pair of data block and tag (m*i ,t *i ) on the server and sends the update message Msginsert to the auditor. Step 3: Delete(mi ) → Msgdelete The deletion algorithm takes as input the data block mi . It outputs the update message Msgdelete  (i, B*i ,V *i ,T *i ). It then deletes the pair of data block and its tag (m*i ,t *i ) from the server and sends the update message Msgdelete to the auditor (Fig. 7). Dynamic Index Update (phase 2): Upon receiving the three types of update messages, the auditor calls three corresponding algorithms to update the ITable. Each algorithm is designed as follows:

52

R. Kumar and U. K. Wiil

IModify (Msgmodify ). The index modification algorithm takes the update message Msgmodify as input. It replaces the version number Vi by the new one V*i and modifies Ti by the new time stamp T *i . IInsert (Msginsert ). The index insertion algorithm takes as input the update message Msginsert . It inserts a new record (i, B*i ,V *i ,T *i ) in ith position in the ITable. It then moves the original ith record and other records after the ith position in the previous ITable backward in order, with the index number increased by 1. IDelete(Msgdelete ). The index deletion algorithm takes as input the update message Msgdelete . It deletes the ith record (i, B*i ,V *i ,T *i ) in the ITable and all the records after the ith position in the original ITable moved forward in order, with the index number decreased by 1. Confirmation Auditing: In our auditing construction, the auditing protocol only involves two-way communication: Challenge and Proof. During the confirmation auditing phase, the owner requires the auditor to check whether the owner’s data are correctly stored on the server. The auditor conducts the confirmation auditing phase as 1. The auditor runs the challenge algorithm Chall to generate the challenge C for all the data blocks in the data component and sends the C  ({i,vi }iQ, R) to the server. 2. Upon receiving the challenge C from the auditor, the server runs the prove algorithm Prove to generate the proof P  (TP,TD) and sends it back to the auditor. 3. When auditor receives the proof P from the server, it runs the verification algorithm Verify to check the correctness of P and extract the auditing result. The auditor then sends the auditing result to the owner. The auditor then sends the auditing result to the owner. If the result is true, the owner is convinced that its data are correctly stored on the server, and it may choose to delete the local version of the data. If the result is true, the owner is convinced that its data are correctly stored on the server, and it may choose to delete the local version of the data. Static Auditing confirmation (Fig. 8). Dynamic Auditing Confirmation (Fig. 9).

Fig. 7 Dynamic index update (Phase 2)

Design and Analysis of an Enhanced Multifactor …

53

Fig. 8 Reference [1] Phase 3 (static confirmation auditing)

Fig. 9 Dynamic update confirmation (Phase 3)

Update confirmation: After the auditor updates the ITable, it conducts a confirmation auditing for the updated data and sends the result to the owner. Then, the owner can choose to delete the local version of data according to the update confirmation auditing result. The Index signifies the current block number of data block mi in the data component M.Dat tags are produced based on time stamps provided by solution. This ITable is created by the owner during the owner initialization and managed by the auditor. When the owner completes the data dynamic operations, it sends an update message to the auditor for updating the ITable that is stored on the auditor. After the confirmation auditing, the auditor sends the result to the owner for the confirmation

54

R. Kumar and U. K. Wiil

that the owner’s data on the server and the abstraction information on the auditor are both up to date. This completes the data dynamic operation. This work is very much prevent the attack but a slighter limitation is the storage while performing operations which are critical so in this research will purpose a lighter detect based TagGen and ITable process which will be carry in packet header itself with 2 bit of information. For the proposed work, focus will be on pseudonym based auditing and saving auditing storage against different attacks. In auditing process, the auditing protocol only involves two-way communications: Challenge and Proof. During the confirmation auditing phase, the owner requires the auditor to check whether the owner’s data are correctly stored on the server. Further auditing data will be updated according to the required storage solutions. The auditor then sends the auditing result to the owner. If the result is true, the owner is convinced that its data are correctly stored on the server, and it may choose to delete the local version of the data. The pseudonyms will compose of public key, private key, and a certificate will be used for efficient preservation of privacy. Users can be assured of their anonymity through pseudonym and authenticated as normal users through a certificate. The TTP stores pseudonyms and actual ID of users to reveal the anonymity in case of a problem and it is also used for storing results for whole auditing process. Then process the operations based on the traces like ITable generation and TagGen generation but storing the trace initial data in header so that it will not create overhead to running processes. Network Simulator will be used for experimentation with dense network. Cloud server structure will be opted for experimentation with various users [13, 14]. Table 2 illustrates a comparison between previous and proposed work. We compare the techniques in [1–9, 11, 12] against the work proposed in this article. It is found that the proposed enhanced multifactor authentication through a covert approach is impregnable and efficient against different attacks.

4 Results The results of the analysis of multifactor authentication through a covert approach are as follows. (a) Communication overhead Communication cost is an excess overhead of computation time, memory, bandwidth, etc. Figure 10 illustrates the communication overhead versus attribute vector dimension for multifactor authentication through a covert approach. (b) Time cost of individual client The time cost tradeoff problem is an important aspect in the scheduling of real time industrial projects. Figure 11 illustrates the time cost of individual client for multifactor authentication through a covert approach. (c) Key generation time taken analysis

[1]

Yes

No

No

No

No

Details

Constant bandwidth cost

Protecting data privacy

Batch auditing

Data owner

Data dynamic support

Yes

No

No

No

No

[2]

Yes

No

No

Yes

No

[3]

Yes

No

No

No

No

[4]

Table 2 A comparison on different dynamic auditing protocols

No

No

No

No

Yes

[5]

Yes

No

No

No

No

[6]

No

Yes

No

No

Yes

[7]

Yes

No

No

No

No

[2]

Yes

No

No

No

No

[7]

No

No

No

No

Yes

[8]

Yes

No

No

No

No

[9]

No

No

No

No

Yes

[11]

No

Yes

No

No

Yes

[12]

Yes

Yes

Yes

Yes

Yes

Proposed

Design and Analysis of an Enhanced Multifactor … 55

56

R. Kumar and U. K. Wiil

Fig. 10 Communication overhead versus attribute vector dimension for multifactor authentication through a covert approach

Fig. 11 Time cost of individual client for multifactor authentication through a covert approach

It is the time taken for key generation. Figure 12 illustrates the key generation time taken analysis for multifactor authentication through a covert approach. (d) Throughput The network throughput of a connection with flow control, for example a TCP connection, with a certain window size (buffer size). Figure 13 illustrates throughput for multifactor authentication through a covert approach. (e) Entropy Here value of entropy is calculated to measure the tendency of a process, to proceed in a direction. Moreover, entropy lends an expression for an encryption method. We

Design and Analysis of an Enhanced Multifactor …

57

Fig. 12 Key generation time taken analysis for multifactor authentication through a covert approach

Fig. 13 Throughput for multifactor authentication through a covert approach

evaluate our hypothesis based on entropy generated. The Fig. 14 depicts entropy for multifactor authentication through a covert approach. (f) Intuitive synthesis It shows improvements in time, space complexity and communication overhead. Here the evaluation of floating frequency of proposed optimized dynamic auditing protocol. Figure 15 shows the intuitive synthesis for multifactor authentication through a covert approach. (g) ASCII histogram

58

R. Kumar and U. K. Wiil

Fig. 14 Entropy for multifactor authentication through a covert approach

Fig. 15 Intuitive synthesis for multifactor authentication through a covert approach

Here with an ASCII histogram, we show the probability calculations based on simple print statements. The calculation of probabilistic simulations is very difficult because results of the given operations will never be the same. However, they should have the same probability distribution, so by looking the shape of the histogram one can understand that calculations are going in right direction. In this context, an ASCII histogram is computed for enhanced secure dynamic auditing protocol. Figure 16 displays that ASCII Histogram for multifactor authentication through a covert approach. (h) Lagged correlation Here the mathematical representation of degree of resemblance between the given and lagged time over consecutive is shown. Figure 17 shows the lagged correlation for multifactor authentication through a covert approach. Some kind of summary of the above analysis would be good. How does it demonstrate the “power” of the proposed solution?

Design and Analysis of an Enhanced Multifactor …

59

Fig. 16 ASCII histogram for multifactor authentication through a covert approach

Fig. 17 Lagged correlation for multifactor authentication through a covert approach

Various aspects of the proposed work is examined including communication overhead, time cost of individual client, key generation time taken analysis, throughput, entropy, intuitive synthesis, ASCII histogram and lagged correlation with added dimensions static and dynamic auditing confirmation. The analysis demonstrates that the proposed scheme is secure and efficient against notorious conspiracy attacks

60

R. Kumar and U. K. Wiil

5 Conclusion Current solutions must continue to evolve and improve their authentication of users moving beyond the limitations of one factor, two factor, etc. Multifactor authentication is far better, if we use this schema both statically and dynamically. These days contextual authentication is seen as complementary to active and explicit authentication factors, but in the future, multifactor authentication prove more secure for the optimal combination of cost-effectiveness, usability, security, etc. We process the operations based on the traces generation but storing the trace initial data in the header so that it will not create overhead to running processes. In this paper, we have proposed an enhanced multifactor authentication through a covert approach and have examined communication overhead, time cost of individual client, key generation time taken analysis, throughput, entropy, intuitive synthesis, ASCII histogram and lagged correlation with added dimensions static and dynamic auditing confirmation. The proposed work may be useful various applications as listed below: • • • • • • • •

Symantec’s Validation and ID Protection (VIP) Service Vasco’s IDENTIKEY Server v3.6 Dell Defender Okta Verify SecureAuth IdP v8.0 CA Strong Authentication SafeNet Authentication Service EMC RSA Authentication Manager and SecurID

References 1. Hwang, M.: Cryptanalysis of remote login authentication scheme. Comput. J. Commun. 22(8), 742–744 (1999) 2. Hwang, M., Chong, S., Chen, T.: DoS- resistant ID-based password authentication scheme using smart cards. Comput. J. Syst. Softw. 83(1), 163–172 (2010) 3. Hwang, M., Lee, C., Tang, Y.: An improvement of SPLICE/AS in WIDE against guessing attack. Internet J. Inf. 12(2), 297–302 (2001) 4. Scott, M.: Cryptanalysis of an Id-based password authentication scheme using smart cards and fingerprints. Comput. J. SIGOPS Oper. Syst. Rev. 38(2), 73–75 (2004) 5. Wang, B., Li, J., Tong, Z.: Cryptanalysis of an enhanced timestamp-based password authentication scheme. Comput. J. Secur. 22(7), 643–645 (2003) 6. Yang, G., Wonga, D., Wang, H., Deng, X.: Two-factor mutual authentication based on smart cards and passwords. J. Comput. Syst. Sci. 74(7), 1160–1172 (2008) 7. Lee, Y., Kim, S., Won, D.: Enhancement of two-factor authenticated key exchange protocols in public wireless LANs. Comput. Electr. Eng. 36(1), 213–223 (2010) 8. Wu, S., Zhu, Y.: Improved two-factor authenticated key exchange protocol. Int. Arab. J. Inf. Technol. 8(4) (2011) 9. Das, M.L.: Two-factor user authentication in wireless sensor networks. IEEE Trans. Wirel. Commun. 8, 1086–1090 (2009)

Design and Analysis of an Enhanced Multifactor …

61

10. Khan, M.K., Alghathbar, K.: Cryptanalysis and security improvements of ‘two-factor user authentication in wireless sensor networks’. Sensors 10, 2450–2459 (2010) 11. Nyang, D.H., Lee, M.K.: Improvement of Das’s two-factor authentication protocol in wireless sensor networks, cryptology ePrint Archive 2009/631. Online PDF. http://eprint.iacr.org/2009/ 631.pdf. Accessed on 28 Feb 2010 12. Pu, Q.: An improved two-factor authentication protocol. In: Second International Conference on MultiMedia and Information Technology (2010) 13. Kumar, R., Verma, H.K., Dhir, R.: Analysis and design of protocol for enhanced threshold proxy signature scheme based on RSA for known signers. Wirel. Pers. Commun. Int. J. (Springer) 80(3), pp. 1281–1345 (2015). ISSN 0929-6212 (Print) 1572-834X (Online) 14. Kumar, R.: Cryptanalysis of protocol for enhanced threshold proxy signature scheme based on elliptic curve cryptography for known signers. In: Knowledge Computing and Its Applications. Springer Nature Singapore Pte Ltd. (2018). https://doi.org/10.1007/978, (ISBN 978-981-106679-5)

Review on Current Trends of Deep Learning Stuti Mehla, Anjali Chaudhary and Raman Kumar

Abstract Artificial Intelligence is that term which take Science to new horizons. In field of computers, AI is described as where machine acts like human. AI has different fields on the basis of problem i.e. machine learning, natural language processing, computer vision and robotics. To achieve these objectives several approaches are in trend like Symbolic Reasoning, Neural Network, Deep Learning and Evolutionary Algorithms. Out of these, Neural Network and Deep learning are those approaches which attracts the researchers. Both are inspired by biological neural network but Deep learning is more refined neural network in which feature extraction and abstraction is automatic as compared to Neural Network. In this chapter we will emphasise on AI technologies and then focus on recent researches in field of Deep Learning i.e. Sentiment Analysis, WSN etc. Keywords AI · Symbolic reasoning · RNN · Sentiment analysis · WSN

1 Introduction According to Smitther Artificial Intelligence is that field of computer science where intelligent behavior is created artificially. AI focuses on to make such machines which act intelligent as human mind do. Basic aim of AI is to make cognitive machines which can find solutions of problem solving tasks, logical reasoning and can do S. Mehla (B) Maharishi Markandeshwar Deemed University, Mullana, Ambala, India e-mail: [email protected] A. Chaudhary Panipat Institute of Engineering and Technology, Panipat, Haryana, India e-mail: [email protected] R. Kumar Department of Computer Science and Engineering, I K Gujral Punjab Technical University, Kapurthala, Punjab, India e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_4

63

64

S. Mehla et al.

Fig. 1 Types of machine learning

Machine learning

Supervised learning

Unsupervised learning

learning by thinking, deducing and learning as human mind do.AI has two types of learning i.e. Symbolic learning and Machine learning. Symbolic learning problems are based on logic and search. This field of AI based on problems, logic and search. Whereas, machine learning is that field when we are making our machine intelligent as human mind. Machine learning is further categorized into natural language processing, Speech and image recognition and expert system. Machine learning is divided into two categories supervised learning and unsupervised learning. In supervised learning program is trained with predefined example to get the desired output and in Unsupervised learning program is developed in which pattern are predicted from a set of data (Fig. 1). Hierarchial tree of AI is explained in which learning is explained. Symbolic learning is defined as and these learnings are categorized further according to problems. The intelligence which is shown by machines according to the principle of natural intelligence in human known as Artificial Intelligence. Devices which perceive according to environment and take action cognitively known as agents. There are two types of learning Symbolic learning and Machine learning. In Symbolic Learning all information is represented in form of rules and the data is to find out through deduction like in Expert System. In contrast Machine Learning is that technology which works on principle of learning of statistical techniques. Machine Learning focus on algorithms that work on prediction by developing a sample model of inputs. This Fig. 2 explains how Artificial Intelligence field has shown a high growth phase in research area of Machine Learning methodologies like Deep Learning (DL) and Reinforcement Learning (RL). Combinations of those techniques determine remarkable performance in solving a wide range of problems. Machine learning is achieved by symbolic reasoning, neural network and evolutionary algorithms. Symbolic learning has given birth to different fields such as robotics, image processing and computer vision. Robotics is that field of engineering in which machines are developed like as human activities. It is interdisciplinary research area including mechanical, electronics and computer science. Images are good source of information through which required data can be fetch out. Process of extracting data from source image is the basic principle of Image Processing. Build up noise, signal distortion are the challenges in image processing. Statistical learning is that learn-

Review on Current Trends of Deep Learning

65 AI

Symbolic Learning

Robotics

Image Processing

Machine learning

Computer vision

Deep Learning

Statistical Learning

CNN Speech Recoginition

RNN

NLP

Fig. 2 Hierarchial tree of AI

ing which deals on prediction of data. Statistical techniques are used to implement prediction. This learning can be used in image processing, computer vision. On the another hand machine learning mimics the human nature i.e. to find the best solution of problem from different solutions. Artificial neural network are just like biological neural networks. In ANN different units are connected like neurons in brain. Connection between units are called edges and these edges are having weight. Weight is adjusted i.e. positive or negative will be on strength of connection. In ANN there is one input layer, one output layer and rest are intermediate layer in which processing is done. Basic aim of ANN is to work like human brain but there is drawback in this we can’t predict with it i.e. overcomed by Deep learning. In deep learning feature extraction and abstraction is done automatically. This relationship can be understanded by venn diagram which

66

S. Mehla et al.

Machine Learning

ANN

Deep Learning

Fig. 3 Relationship between Machine learning, ANN and Deep Learning

is shown in Fig. 3. Deep learning predict the latest solutions, so it has become the latest topic of research for academicians, companies. Deep learning algorithms are used to find the unknown structure to find out the high level feature extraction. On the basis of extraction Sentiment analysis is basically of two types: knowledge and statistics dependable. Knowledge based sentiment analysis worked on polarity of data while Statistics based sentiment analysis make a target on supervisal statistical techniques [1]. Main objective is to make higher level feature extraction more fruitful. In this paper unsupervised pretraining approach is implemented and former hypothesis of probability is also enhanced [2]. In another research work a new model is developed in initialization phase which automatically discard the extra features. In this research work unsupervised neural network is developed in which during initialization phase neural network is embedded with deep learning model. Output stage will contain these pre trained data set [3]. As the social media has given birth to sentiment analysis. Sentiment analysis is defined to predict the polarity from given sample of views but subjective interpretation and Linguistic phenomenon are another drawbacks which are overcomed by deep learning [4]. In this research work meaningful features are extracted from every review. Practical work is carried out on 22 data sets of different domains [5]. As sentiment analysis of short texts is very difficult because prediction is very difficult in short texts so a deep convolution neural network is developed to back this problem and the experiments are carried out on two domains: one is twitter data related to bank and another is related to movies [3]. A supervised framework for sentiment analysis is also developed termed as Coooolll to extract the features from twitter data. In this research work a new concept of hybrid loss work is introduced. Practicals are carried on sample data and it is valid for real time data also [6]. There is a difference of common sense between human being and machine. It is really difficult to take decision with senses to achieve them SenticNet was proposed. But SenticNet can’t generalize this knowledge and it is overcome by SenticNet 4. In this research work conceptual perception is foundation. SenticNet 4 work on the principle of similarity proposed by Gestalt. In this framework automatic conceptual faults are resolved while performing hierarchial clustering and dimensionality reduction [1]. In unsupervised deep learning, to get better results author proposed a

Review on Current Trends of Deep Learning

67

model which focuses on word embeddings during the process of initialization and then it is combined with deep learning concepts [7]. Another field in which deep learning is expanding i.e. IOT. According to Gartner Internet of things is defined as network of physical objects containing embedded technology to sense and interact with external environment. There are some terms in IOT which are explained below. Electronic environment which is sensitive and responds according to presence of people by taking consideration of their daily activities. and rituals known as Ambient Intelligence [2]. Smart object is defined as that object which enhances the interaction with people as well as smart things. It can interact with not only physical objects but also virtual environment. A case study is explained below regarding relationship between deep learning and IOT [4]. In this case study, an Open Source machine learning platform for smarter applications is introduced. In this case study, we have been following H2 O for features such as Open Source, R integration, Spark integration, Deep Learning [3]. Internet of Things (IoT) is a new revolution of the Internet as it makes Objects themselves recognizable, obtain intelligence, communicate information about themselves and they can access information that has been aggregated by other things. The Internet of Things allows people and things to be connected Anytime, Anyplace, with any thing and anyone, ideally using any path/network and any service. This implies addressing elements such as Convergence, Content, Collections, Computing, Communication, and Connectivity. The Internet of Things provides interaction among the real/physical and the digital/virtual worlds. The physical entities have digital counterparts and virtual representation and things. Will be more context aware and they can sense, communicate, interact, exchange data, information and knowledge. Through the use of intelligent decision-making algorithms in software applications, appropriate rapid responses can be given to physical entity based on the very latest information collected about physical entities and consideration of patterns in the historical data, either for the same entity or for similar entities. These paves new dimension of IoT concept in the domains such as supply chain management, transportation and logistics, aerospace, and automotive, smart environments (homes, buildings, infrastructure), energy, defence, agriculture, retail and more. The vision of IoT is to use smart technologies to connect things any-time, any-place. The IoT was started in the year 1998 and their everyday lives term IOT was first coined by Kevin Ashton in 1999. In last two decades, the Internet of things(IOT) solutions starts to emerges from the initial pioneering visions to regular industrial solutions, which are present in our everyday lives so because of this a number of challenges get generated, so the contents specify the quality assurance and testing techniques for IOT domain. Analytics are at the core of IoT. To implement IoT analytics, we need to apply Machine Learning Algorithms to IoT datasets. However, the methodology for machine learning implementations is different from traditional techniques. Several

68

S. Mehla et al.

factors and constraints apply when we consider Data Science for IoT implementations. These include: • Data capture frequency: Data is not produced at same frequency in most of the devices and information systems. • Variety of data sources: To produce right analytics, data from different sources which includes historical and real time systems need to be brought together. • Minimal human intervention: Typically, Internet of things implies minimal Human intervention. IoT analytics algorithms need to cater for this requirement. Many features of H2 O provide the Smart capabilities for IoT Analytics—for instance In-Memory processing, the Deep learning Package and the Spark streaming package (Sparking Water). The Deep learning package from H2 O features automatic data handling techniques to standardize data, handle missing values and categorical data conversions. It also includes automatic performance tuning, load balancing and it takes out the overhead of complicated configuration files. First, we look into the deep learning capabilities of H2 O. This will provide a backdrop for the use of H2 O for building IOT capabilities. In a nutshell, Deep learning methods implements neural network algorithms such as Feed Forward Back Propagation Neural Network, Convoluted Neural Networks, Recursive Neural Networks and others. Deep learning algorithms play an important role in IOT analytics. Data from machines is sparse and/or has a temporal element in it. Even when we trust data from a specific device, devices may behave differently at different conditions. Hence, capturing all scenarios for data preprocessing/training stage of an algorithm is difficult. Monitoring sensor data continuously is also cumbersome and expensive. Deep learning algorithms can help to mitigate these risks. Deep Learning algorithms learn on their own allowing the developer to concentrate on better things without worrying about training them. Below, we discuss what H2 O offers as a part of its deep learning framework and features that makes it suitable for data from things. Some useful examples of IOT are explained below. Automated vacuum cleaners, like that of the iRobot Roomba i.e. an example of IOT. iRobot set the standard with it’s first commercially successful automated vacuum in 2002. Founded by MIT roboticist, the company has developed technology to help it’s puck-shaped vacuum robots to map and “remember” a home layout, adapt to different surfaces or new items, clean a room with the most efficient movement pattern, and dock itself to recharge it’s batteries. Though the “smart home” hasn’t exactly revolutionized life for most of us, some companies are ardently aiming to change that—and there’s few better examples than Nest, the company acquired by Google for a reported $3.2 billion. As an IoT device, Nest’s clean digital interface is (for many) a welcome change from the clunkier physical dial, and it’s smartphone integration allow for temperature checking and controls from anywhere. Self-driving vehicles, such as that of Tesla Motors is also another example. Cars are “things,” and in so much as we’re interested in “things” that leverage powerful artificial intelligence, automotive technology is ahead of the curve. This isn’t necessarily because autonomous vehicles will be the

Review on Current Trends of Deep Learning

69

easiest IoT innovation to bring to life, but with nearly all major car manufacturers throwing billions of dollars at the problem, it certainly has momentum To use Tesla’s technology as an example, we’ll need to understand how Tesla’s autonomous vehicle technology really works. In an IoT Network, the primary technologies are wirelesspersonal area network viz. 6LoWPAN, ZigBee, Bluetooth. In enlargement, on a slightly larger wireless network scale, Wi-Fi, Wireless LAN technology, is to be used for supporting, so According to the need of larger scale network, the backbone is used in addition to the mobile communication network domain. Then smartphones and mobile communication phones will also be used because of connection to the base stations. The base stations basically used to provide connectivity to the wider network, the internet. According to current technologies requirements smartphones are equipped with Bluetooth and Wi-Fi because of what it’s possible to establish IoTNetwork. The same as Bluetooth and some other type of network provide a simple topology control using wireless Pan which is better from each constraint. The process defines that how the smartphone, carry that signal up and connect it through LTE, or 4G/3G to the base station and the base station will connect to that widearea network, which is the internet. The observation states that one technology is linking on a technology of something. While considering wearable IoT Networks devices such as shoes, watch, glasses, belt, etc. can be used to detect biometric information because they are close and attached to body and they can sense in real time. These devices can send data through Bluetooth technology to the smart phone that rests inside the pocket. In that case, the smartphone will be able to pick up this information and it will be able to use it through mobile communication link to send it to a base station. The base station, which is connected to a wide area network and the Internet, is going to be where the information connects to any other place we need in the world. For example, if you are using your watch and also your shoes as well as your belt or something similar you have attached to self, to monitor your health condition that can be sent wirelessly to your smartphone then, your smartphone will send that information to the base station, which will be connected to the internet and that information can be delivered to a control center or medical server. If something happens to you, they will know and they will be able to provide immediate support. A smart device can be used to collect the information that communicates with control centers or like a medical center and this is how everything gets control. A number of discussions have been conducted regarding the IoT issues for however, during our literature survey, we have not found a systematic analysis, identifying what is the impact of these specifics to particular software testing methods and techniques. Hence, we provide such an analysis in this paper. In the following section, we identify several typical issues of IOT solutions and we number them by IDs.

70

S. Mehla et al.

2 Conclusion Deep Learning is emerged as that technology which can be used for different purposes like in sentiment analysis, image processing, WSNs and IOT. It can be further used in Big Data Analytics. The basic idea of finding the solution from huge amount of data using intuitive approach like human make it useful for every field.

References 1. Kumar, R., Verma, H.K., Dhir, R.: Analysis and design of protocol for enhanced threshold proxy signature scheme based on RSA for known signers. Wirel. Personal Commun.—Int. J. 80(3), 1281–1345 (2015). ISSN 0929-6212 (Print) 1572-834X (Online) 2. Bengio, Y.: Deep Learning of Representations for Unsupervised and Transfer Learning 3. Severyn, A., Moschitti, A.: Twitter Sentiment Analysis with Deep Convolutional Neural Networks 4. Rojas Barhona, L.M.: Deep Learning for Sentiment Analysis 5. Glorot, X., Bordes, A., Bengio, Y.: Domain Adaptation for Large-Scale Sentiment Classification: A Deep Learning Approach 6. https://www.kdnuggets.com/2016/04/deep-learning-iot-h2o.html 7. Kumar, R.: Cryptanalysis of protocol for enhanced threshold proxy signature scheme based on elliptic curve cryptography for known signers. Knowl. Comput. Appl. (2018). https://doi.org/ 10.1007/978. ISBN 978-981-10-6679-5. Springer Nature Singapore Pte Ltd. 8. dos Santos, C.N., Gatti, M.: Deep Convolutional Neural Networks for Sentiment Analysis of Short Texts 9. https://medium.com/iotforall/using-deep-learning-processors-for-intelligent-iot-devices1a7ed9d2226d 10. https://en.wikipedia.org/wiki/Smart_objects

Hybrid Homomorphic Encryption Scheme for Secure Cloud Data Storage Bijeta Seth, Surjeet Dalal and Raman Kumar

Abstract Cloud computing is the budding paradigm nowadays in the macrocosm of the information processing system. It offers a variety of services for the users through the Internet and is highly cost-efficient and flexible. Information storage in the cloud is showing great attention. Yet, despite all its advantages, security and privacy has evolved to be of significant apprehension in cloud computing and is discouraging factor for potential adopters. Consumers and businesses prefer online computing only if their data are guaranteed to remain secret and safe. Hence, the focal point is to discover techniques in the direction of offering more confidentiality. Homomorphic encryption is one such technique. This paper targets to study several key concepts of cloud computing, namely characteristics, delivery models, deployment models and cloud computing platforms. The theme includes the security challenges/issues and their associated work in cloud computing. A generic flow of data and the operations to be performed in the proposed scheme for multi-clouds are presented. The report explains the details and effects related to different parameters of Homomorphic properties of some cryptosystems. In this paper, our main work is to ensure the protection of information, thus we offered a method to amend the Paillier Homomorphic algorithm without compromising the protection of existing technique. In our prospect work, we plan to propose an efficient Multicloud architecture so that information is stored, maintained and retrieved efficiently by employing a modified Paillier approach. Keywords Cloud computing · Security · Issues · Attacks · Homomorphic encryption · RSA · ElGamal · Paillier · Encryption · Decryption · Entropy B. Seth (B) · S. Dalal SRM University, Sonepat, Haryana, India e-mail: [email protected] S. Dalal e-mail: [email protected] R. Kumar Department of Computer Science and Engineering, I K Gujral Punjab Technical University, Kapurthala, Punjab, India e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_5

71

72

B. Seth et al.

1 Introduction In 2008, Cloud computing has evolved as a new revolution in Information technology as it provides a variety of services and applications to be run from anywhere in the world to its users through the Internet. Most of the operations involve a trusted third party. The cloud should trust an entity, human or machine to preserve the confidentiality of the data. But an attack on the trusted party could reveal all the sensitive data, therefore the requirement where even the service providers have no information about users’ data is growing. Homomorphic encryption is one such method. Homomorphic crypto-systems are emerging to be extremely beneficial and exciting; nonetheless, there is even a large amount of research that needs to be exercised to get these systems to be made practical with benefits. The clause is structured as follows: Sect. 1 discusses cloud characteristics, delivery models and service offerings and cloud platform and technologies. Section 2 mentions security in Clouds discussing challenges and issues. Part 3 describes realworld case studies. Part 4 highlights the survey made out in the frame of a table and Sect. 5 describes Homomorphic encryption, its types, flavors, benefits, and restrictions. Further, Sect. 6 mentions an overview of discussing Homomorphic algorithms. Section 7 describes the proposed algorithm and operations to be performed on files in multiclouds. Finally, Results and Discussion constitute Sect. 8. Conclusion and Future scope comprise the last part of the paper.

1.1 Cloud Characteristics Cloud computing has namely, following subsequent crucial characteristics as described by The Cloud Security Alliance [1] like Self-service as per requirement, Broad Internet access, the Huge pool of Resources, Rapid elasticity and Service as per usage (Fig. 1).

1.2 Delivery Models Cloud services can be provided as four basic cloud delivery models, namely Public cloud which provides the interface between the unrestricted customers and the owner group (third party), for example, Amazon cloud service. It is more cost-effective, highly reliable and flexible and location Independent. But they are less secure and customizable. Private cloud affords the services merely for an organization in an exclusive manner. For example, CIO /G6. It provides high security and more control

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

73

Fig. 1 Cloud computing paradigm. Source Abdullah [2]

in comparison of public clouds. But such models have restricted area of operation and have a high price with limited scalability. Community cloud provides the services for the specific groups instead of whole public groups. They all solve together for common businesses. For example, Government or G-Cloud. Cost-effectiveness and more security are the vantages of using community clouds. Hybrid cloud is formed by combining any of the public, private or community clouds. For example, CIO/G6/APC+ Amazon EC2. Enhanced scalability, security, and flexibility are the advantages of this model. But it faces networking issues and security compliance.

1.3 Cloud Service Offerings The underlying character of cloud service offerings is Software as a service (SaaS) which permits the consumer to include an application for lease from cloud service provider instead of buying, establishing and operating software. For Example, Gmail Docs. Platform as a service (PaaS) cloud which gives a provision to the users upon which applications can be coordinated and carried out. For Example, Windows Azure. Infrastructure as a service (IaaS) where the users are capable of accessing resources according to their requirements for huge pools installed in data centers. For example, Elastic Cloud Compute.

74

B. Seth et al.

1.4 Cloud Computing Platforms and Technologies Here we discuss different platforms and frameworks in cloud computing like Amazon Web Services (AWS) is a group of web services that work in cooperation to deliver cloud services. It allows users to store and copy information across geographical areas. Google App Engine is an Internet-based aggregation of applications which use distributed file system (DFS) to gather information. It offers single sign-on (SSO) service to integrate with LDAP. Microsoft Azure structures the applications around the notion of roles. Azure recognizes and embodies a distribution unit for an application mainly web role, worker role, and virtual machine role. Hadoop Apache which is an open source structure and is desirable for processing large data sets on commodity hardware. Hadoop is an implementation of MapReduce, which provides two fundamental operations for data processing: map and reduce. Salesforce provides Force.com for building business applications and uses Stateful packet inspection firewall. For authentication purposes, LDAP is used. Unknown address connection requests are refused.

2 Security in Cloud Environment This section cites the challenges and issues connected with cloud computing.

2.1 Challenges Following are the principal challenges that occur in the adoption of clouds [3]: • Outsourcing: Privacy violations can occur as the customers actually lose control on their data and tasks. • Multi-tenancy: New vulnerabilities and security issues can happen because of the shared nature of clouds between multiple clients. • Massive Data and Intense Computation: Traditional security mechanisms can’t be applied to clouds due to large computation or communication overhead. • Heterogeneity: Integration problems arise between diverse cloud providers using different security and privacy methods. • Service Level Agreement: A negotiation mechanism between provider and consumer of services need to be established. Security is regarded as the dominant barrier amongst the nine challenges in accordance with the survey done by IDC [4] as shown in Table 1.

Hybrid Homomorphic Encryption Scheme for Secure Cloud … Table 1 Challenges/issues in clouds (results of IDC ranking security challenges (3Q2009, n  263)

75

S. no.

Challenge/issue

%age

1

Security

87.5

2

Performance

83.3

3

Availability

60.1

4

Hard to integrate with in-house IT

61.1

5

Not enough ability to customize

55.8

6

Worried on demand will cost more

50.4

7

Bringing back in-house may be difficult

50.0

8

Regulatory requirements prohibit cloud

49.2

9

Not enough major suppliers yet

44.3

2.2 Cloud Computing Security Issues There are subsequent security issues as designated below [3]: • Trust: The cloud service provider is needed to offer sufficient security policy to lessen the threat of information loss or data manipulation. • Confidentiality: The confidentiality can be breached as sharing or storage of information on remote servers is done in cloud computing which is accessed through the internet. • Privacy: is defined as the readiness of a client to have power over the revelation of private information. An illegal admittance to user’s sensitive data will bring security issues. • Integrity: is to guarantee the precision and uniformity of data. Thence, the Cloud service provider should offer protection against insider attacks on data. • Reliability and Availability: Trustworthiness of cloud service provider decreases when a user’s data get leaked. • Authentication and Authorization [5]: To prevent unauthorized access, software is required outside the organization’s firewall. • Data Loss: Removal or modification of data lacking any backup could lead to data loss. • Easy Accessibility of Cloud: Cloud services are able to be applied by anybody by a simple registration model. This opens a chance to access services for the crafty minds.

3 Case Studies Many real-world scenarios where cloud computing was compromised by attacks and their feasible prevention methods are listed below in Table 2.

76

B. Seth et al.

Table 2 Case Studies Type of attack

Definition

Example

Solution

XML Signature Wrapping Attack

Wrapping attack inserts a fake element into the signature and then makes a web service request

In 2011, Dr. Jorg Schwenk discovered a cryptographic hole in Amazon EC2 and S3 services

A suggested answer is to apply a redundant bit called STAMP bit in the signature in the SOAP message

Malware Injection

Hacker tries to enter malicious code by inserting code, scripts, etc. into a scheme

In May 2009, four public websites were set offline for the BEP in which hackers introduced undetectable iFrame HTML code that redirected guests to a Ukrainian website

WWW browsers like Firefox should install NoScript and set Plugins The FAT table can be applied to ascertain the validity and unity of the new case

Social Engineering Attack

It depends on human interaction, thereby breaking normal security operations

On August 2012, hackers completely destroyed Mat Honan’s digital life by deleting data from his iPad, iPod, and MacBook by exploiting Amazon and AppleID Account of the victim

Apple forced its customers to use Apple’s online “iForgot” system to provide stronger authentication. Various account settings like a credit card, email addresses can’t be altered on phone by Amazon customer service head

Account Hijacking

It compromises confidentiality, integrity, and availability by stealing credentials of accounts

On July 2012, UGNazi, enter CloudFare’s personal Gmail by exploiting Google’s email &password recovery system

CloudFlare has stopped sending password reset and transactional messages for protection use

4 Related Work One of the most complex aims in cloud computing is to provide security and protecting data privacy [6]. Simply due to its shared nature, it gets hard to prevent threats [7] in cloud computing, then data can be leaked by unauthorized access. This section presents an outline of existing review articles allied to security and privacy. Amato et al. [8] illustrated a new approach using Model-driven Engineering techniques for security analysis [9] and monitoring of cloud infrastructures by providing thermal based behavior of hosts. Cloud is

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

77

supposed to be a compulsory prerequisite for organizations employing Big-data [10] services. But the security of data exists as a major issue while deploying data on physical hosts which are virtually separated. Mohammed [11] proposed a new paradigm “Mobile Cloud computing” by integrating mobile computing into cloud computing. The communication between the mobile user and clouds held through the wireless transmission medium introduced new security and privacy challenges. The author discussed some of these issues and compared the work done by researchers based on different security and privacy requirements. Shugrue [12] emphasized that companies conducting online business must ensure that their web applications, websites, and APIs are scrutinized from attacks and vulnerabilities to protect data breaches and data loss. Agoulmine et al. [13] described a mode to increase the confidence in buying services through Service Level Agreements (SLA) which furnishes a perfect characterization of the guarantees accessible by the source to the consumers. The paper elaborates the literature study done to enumerate existing solutions and open issues regarding SLAs in cloud computing. Castiglione et al. [14] mentioned that data Outsourcing among geographically distributed data centers under the different legislative framework require ensuring the privacy of user data and transactions. Peng [10] specified cloud storage as an efficient data management technique limited by the security concerns of users about the third-party cloud service providers. The author mentioned Oblivious RAM denoted as ORAM as a technique to facilitate privacy-preserving access to information stored in the cloud and discussed the various load balancing dilemmas when ORAM is applied to Big Data in the cloud. Varadharajan [15] suggested that security techniques must be integrated into a dynamic environment to enhance security for securing distributed applications. The proposed security architecture mentioned intrusion detection and trusted attestation techniques to identify and counter dynamic attacks efficiently and thereby increasing the lifecycle of virtual machines. The resilience of the system is enhanced by employing the opinion amid the various security mechanisms which detect dynamic and sophisticated attacks. Benzeki [16] Mentioned how homomorphic encryption is considered to be appropriate for storing data onto a cloud and mentioned various events connected to it. The author described security and privacy implications, challenges and approaches. Naskos [17] described that Elastic cloud applications have security-related concerns. A solution based on Markov decision process based on probabilistic model checking is proposed illustrating the main steps in online analysis and decision making regarding elastic decisions. Several security aspects were considerably mentioned to be remarkable. Carlson [18] provided a baseline for security analysis of the computational environment considering threats, vulnerabilities, and impacts. The writer noted that the serious threat is non-technical and could be decided by management processes. Rong [19] described Cloud Computing as a model which provides capabilities as “services” in a suitable, on-demand admittance to a collective puddle of assets with nominal organization exertion. The author mentioned that clouds are offset by the apparent security intimidation feared by users and discussed the various security challenges. Service level agreements and

78

B. Seth et al.

holistic mechanisms are viewed to be future solutions for access to clouds. Zang [20] proposed data integrity scheme against malicious auditor and external adversaries by utilizing random masking scheme and Jiawei and Schucheng [21] discussed neural networks and Preservation of privacy was done with multi-layer back propagation neural networks with homomorphic encryption for a multi-party system. Aguiar [22] offered an extensive synopsis of literature covering security aspects of cloud computing, attacks [23] and security mechanisms, maintaining the privacy and integrity of data in cloud computing. Bohli and Nils [24] brought up four major Multicloud approaches, namely application partitioning, tier division, data separation and secure multiparty computation with its drawbacks, compliance with legal obligations, and feasibility. Pearson [25] discussed issues associated to software virtualization, access, to and compliance. The author also mentioned why and how issues like security, trust, and privacy occur in cloud computing. Nepal [26] observed that various workflow environments and tools like Taverna and Galaxy have evolved to be integrated with cloud tools like CloudMap, CloudMan etc. for service delivery and information consolidation. Albeit, Cloud security caused severe challenge concerning the acceptance of cloud-like Health sector where genomics data and medicine sector need much attention. The paper described an architecture termed TruXy for collaborative Bioinformatics research in endocrine genomics. The author compared the performance of TruXy with standalone workflow tool and mentioned its success for handling big data security challenges. Ryan [27] talked about security issues and the advances to handle confidentiality from cloud service providers considering the case of a running SaaS.

5 Homomorphic Encryption The troubles faced by cloud can be figured out by good cloud computing protocols, and as a result, Secure Function Evaluation (SFE) is gaining more significance. SFE provides an important tool when designing protocols where multiple parties exchange their information and still keeping data secret. The development of Homomorphic Encryption is an approach using SFE protocol which can be directly applied to encrypted data. Definition An encryption is Homomorphic if: from En (P) and En (Q) it is possible to compute En (Func(P, Q)), where ‘Func’ can be one of the operations: +, −, * exclusive of using the private key. For instance, adding two encrypted numbers and decoding the end result without being able to recognize the individual worth. Homomorphic encryption was developed in 1978 by Ronald Rivest, Leonard Adleman, and Michael Dertouzos and originated from the concept of privacy homomorphism. Homomorphic encryption (HE) comprises four functions [28], namely Key generation, Encryption, Evaluation and Decryption as shown in the Fig. 2.

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

79

Fig. 2 Homomorphic encryption functions

5.1 Classification of Homomorphic Encryption Multiplicatively Homomorphic: When a permissible action on the encrypted data is constrained to multiplication, it is said to be Multiplicatively Homomorphic. Example: Homomorphic Encryption is multiplicative if Ek (PT1  PT2)  Ek (PT1)  Ek (PT2) Example: RSA Additive Homomorphic Encryption: When an allowable operation on the encrypted data is limited to addition, it is said to be Additively Homomorphic. Example: Ek (PT1  PT2)  Ek (PT1) + Ek (PT2). Example: Paillier

80

B. Seth et al.

Table 3 Comparison of partial and fully HE [30] Parameter

Partial HE

Fully HE

Type of operation supported

It allows either addition or multiplication scheme

It allows both addition and multiplication operations

Computation

It allows a limited number of computations

It allows an unlimited number of computations

Computational efforts

It requires less effort

Requires more efforts

Performance

It is faster and more compact

It has slower performance

Versatility

It is low

It has high

Speed

It is fast in speed

It is slow in speed

Ciphertext size

It is small

It is large

Example

Unpadded RSA, ElGamal

Gentry Scheme

5.2 Types of Homomorphic Encryption Homomorphic encryption is of three types: partial homomorphic system, somewhat homomorphic system and fully homomorphic system [28]. An encryption technique is identified as Somewhat Homomorphic if it performs a restricted amount of addition and multiplication on encrypted information (Table 3).

5.3 Benefits and Limitations of Homomorphic Encryption Homomorphic encryption has several benefits including homomorphic encryption solves the confidentiality problems when data is shared by different users and perform different operations on it, provides privacy by having ability to directly operate on encrypted data, treatment given to patients after analyzing the disease without disclosing the patient details, provides protection of mobile agents and so on. However, its compute and storage overhead has restricted its use.

6 Overview of Discussing Homomorphic Encryption Algorithms In this part, we will briefly introduce three partial homomorphic encryption schemes, namely, RSA [29], ElGamal and Paillier HE schemes [30].

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

81

Fig. 3 RSA algorithm: key generation, encryption, and decryption

6.1 RSA It is the most accepted public key cryptosystem and is extensively meant for digital signatures [31]. It was given by Rivest, Shamir Adleman in 1997 and is shown in Fig. 3. It is multiplicative HE. It provides secure communication. It is utilized for secure internet banking and credit card transaction. Its potency lies on the intractability of an integer factorization dilemma. The security of the scheme lies in the intricacy in factoring n into p and q. To ensure security, the numbers p and q are required to be very large. So far 768-bit RSA has been broken and therefore higher key sizes are suggested for a secure system [32]. It is vulnerable to Brute-Force attack.

6.2 ElGamal It was developed and named after ElGamal [33] is shown in Fig. 4. It is multiplicative in nature. It ensures secure communication and storage. It is widely utilized in hybrid systems. The safety measures of EIGamal method depends on the attributes of the fundamental cyclic group G and padding format used in the messages. The ElGamal scheme is typically utilized in hybrid cryptosystems. Instance, the message being encrypted by the symmetric algorithm and then EIGamal (the asymmetric algorithm is causing the slower speed for the same degree of security) encrypts the key used intended for symmetric encryption. The ElGamal security is dependent on the Discrete Logarithm problem. Choosing large values for prime numbers and random numbers make it difficult to break. The Man in the middle attack can take place because of Forged Signatures being chosen [32].

82

B. Seth et al.

Fig. 4 EIGamal algorithm: key generation, encryption, and decryption

Fig. 5 Paillier algorithm: key generation, encryption, and decryption

6.3 Paillier In 1999, Pascal Paillier developed a probabilistic asymmetric algorithm meant for public key cryptography [34]. It takes in an additive based on “Decimal Composite Residuosity Assumption (DCRA) which makes it intractable. It is additive in nature. It is similar to RSA [35] and uses different keys for encryption and decoding and is shown in Fig. 5. Because of its malleable nature, it is utilized in electronic voting where each vote is encrypted, but only the “sum” is decoded. CryptDB uses Paillier cryptosystem to perform database operations and allows SQL queries to be performed over encrypted data. The Paillier encryption design provides a semantic security against chosen plaintext attacks [36]. The security of the Paillier is depended on Integer Factorization

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

83

where ‘n’ is recommended to be either 2048 or 3072 bits. While selecting the parameter g, g is checked to be multiple of n and it should be taken as small for better performance reasons. Pascal [34]. It can be determined by using the next equation: gcd[L(g λ · modn 2 ), n]  1

(1)

7 Proposed Algorithm In the proposed model, we aim to provide a representation of a special architectural pattern for providing security for multiple cloud providers with the objective to design and develop security mechanism for cloud computing paradigm and to compare the proposed scheme with existing algorithms. The Multi-cloud approach makes use of two or more clouds instead of applying a single cloud and thus avoids reliance on any one single cloud. Why migrate to multi-clouds: • They possess the power to lessen the threat of. – This can result in corruption or loss of data. – Wicked insiders in the lone cloud. • It includes several views of security [37] like confidentiality, integrity, availability, efficient retrieval and information sharing. • It has increased trust and distributed reliability among multiple cloud providers. • It evades “vendor lock-in” by dividing the user’s information between numerous clouds. The modified Paillier having different value of public key, g such as g ∈ [z/n 2 z]such that g λ  1 + n · (mod n2 )

(2)

In the proposed system, the protection of the practice of the public input is presented by modifying the value of public key ‘g’ which is produced from public modulus n and the decryption algorithm. In the decryption process in modified Pail lier, the term in the denominator L g λ · modn 2 Equates to 1. The proposed system is more secure than the original Paillier because g is samplable from z/z 2 in polynomial time of log n and can be generated merely by the public entity, n. The encryption and decryption process of modified Paillier is shown in Fig. 6. The security in multi-clouds can be achieved by using the modified Paillier algorithm. In this paper, we aim to design a framework which allows the users to upload files to a cloud server and a function to split the files into multiple parts by using the concept of crypto-database splitting. Any request for storing the data or retrieving the data from multi-clouds can be fulfilled as a generous flow of data in the proposed scheme as shown in Fig. 7.

84

B. Seth et al.

The client can select the operations of uploading and downloading to be performed using algorithms mentioned below.

In the case of any modification, perform the desired operation and modify all share in different clouds. The security is aimed to be provided using the modified Paillier algorithm which helps to protect data when it is outsourced to the cloud database. So even if the cloud is untrustworthy the data is secured in two ways as the data is split into multiple parts so it is not legible and the data is encrypted using the modified Paillier encryption algorithm making it difficult to decipher it without the key.

Fig. 6 Modified Paillier: key generation, encryption, and decryption

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

85

Fig. 7 Generic multicloud data flow

8 Results and Discussion 8.1 Implementation The module is implemented using JavaNetBeans IDE 8.0.2 on Windows 10 Platform. We compared RSA, ElGamal and Paillier algorithm on the list of parameters like block size, key length, encryption and decryption time and encryption key size and decryption key size to contrast the performance of these algorithms. Encryption time is termed as the time taken to generate ciphertext from the given plaintext of an algorithm. Decryption time is termed as the time taken by the algorithm taken by the algorithm to generate plaintext from the given ciphertext. Figures 8, 9, 10 and 11 shows the file size considered, encryption and decryption time took in milliseconds, encryption and decryption key size of RSA, ElGamal, Paillier and modified Paillier algorithms.

8.2 Discussion and Analysis In the end, the results were obtained which concluded that modified Paillier is considered to be most secure and fastest algorithm. The table below shows the Encryption/Decryption times and key sizes respectively in milliseconds for a file of size 3 bytes as shown in Table 4.

86

B. Seth et al.

Fig. 8 Graph for RSA

The execution time will not depend on the JavaNetBeans IDE 8.02 used as the Executable code is hardware independent. The file size may vary and the performance is not affected by it as we are calculating Encryption and Decryption times along with the key size for various algorithms. • Security Analysis Fine sources of randomness are crucial in cryptography [38], and entropy is frequently employed to measure randomness. Low entropy signifies that the source probably isn’t actually random. In particular, if we have a random variable X that

Table 4 Comparison of various algorithms on basis of different parameters Algorithm

Encryption time (ms)

Decryption time (ms)

Encryption key size

Decryption key size

RSA

10,830

13,054

375

375

EIGamal Paillier Modified Paillier

2264

3361

32

32

21,223

26,627

32

32

331

138

32

32

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

87

Fig. 9 Graph for ElGamal

takes on values x 1 , …, x n with probabilities p(x 1 ), …, p(x n ) respectively, then the entropy of X is  n    (3) H (X )  − p(xi ) log p(xi )  i1

This value is maximized when all of the probabilities are the same. If we have 2n different symbols, that maximum value will be n bits of entropy per symbol. That’s the hypothetical highest level of entropy that we can achieve.

88

B. Seth et al.

Fig. 10 Graph for Paillier

• Numerical Analysis The Security has been analyzed using CrypTool 1.4.30. The Entropy of the mentioned algorithms is considered as an evaluation parameter. The entropy of a document is an index of its information content and is calculated in bits per character. Higher the entropy value signifies higher the hardness of the key. From the evaluation table, we observed that the entropy of proposed Paillier is higher than other algorithms (Table 5). The bar graph for considering parameters among RSA, ElGamal, Paillier and proposed Paillier has been shown in Fig. 12 and for entropy in Fig. 13.

Table 5 Comparison table for entropy

Algorithm

Entropy

RSA

1.5

ElGamal

2.53

Paillier

4.31

Modified Paillier

4.49

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

Fig. 11 Graph for modified Paillier

Fig. 12 Bar graph for various discussed algorithms

89

90

B. Seth et al.

Fig. 13 Bar graph for entropy

9 Conclusion and Potential Work Cloud computing is the most recent growth in online computing. Because storage and computing services are provided in clouds at very low cost, Cloud computing is becoming very popular. The article provided a description of literature covering security aspects of cloud computing. Our knowledge indicates that Security and Privacy are the chief concerns that are compulsory to be considered. This document has addressed several security approaches to overcome the issues in security in cloud computing. Various real-world examples illustrating attacks in cloud computing were discussed. The purpose of the paper was to study and investigate the principal of Homomorphic mechanisms to provide security. The modules were implemented in JavaNetBeans IDE 8.0.2 and Security Analysis was done in CrypTool 1.4.30. For future enhancements, efforts are being made to build up a Multicloud architecture as an efficient scheme that can provide security using Homomorphic schemes taking into account the various security attacks.

References 1. Zhifeng, X.: Security and privacy in cloud computing. IEEE Commun. Surv. Tutorials 15(2), 843–859. (2013). https://doi.org/10.1109/SURV.2012.060912.00182 2. Abdullah, A.: Evaluating security and privacy in cloud computing services: a stakeholder’s perspective. In: The 7th International Conference for Internet Technology and Secured Transactions ICITST IEEE. ISBN 978-1-908320-08/7. (2014) 3. Banafar, H., Sharma, S.: Secure cloud environment using hidden markov model and rule-based generation. Int. J. Comput. Sci. Inf. Technol. 5(3), 4808–4817 (2014) 4. Gens, F.: New IDC IT cloud services survey: top benefits and challenges. IDC. http://blogs. idc.com/ie/?p=730 (2009)

Hybrid Homomorphic Encryption Scheme for Secure Cloud …

91

5. Gajeli, P.: Data security in public cloud for authorization. IJCSE 5(6), 198–202 (2017) 6. Tari, Z.: Security and privacy in cloud computing. IEEE Cloud Comput. 1(1), 54–57. https:// doi.org/10.1109/mcc.2014.20. RMIT University (2014) 7. Sood, R.: Cloud security threats and issues-A review. IJCSE 5(4), 115–119 (2017) 8. Amato, F., Moscato, F., et al.: Improving security in the cloud by formal modeling of IaaS resources. Elsevier. https://doi.org/10.1016/j.future.2017.08.016 (2017) 9. Carlson, R.: Frederick, security analysis of cloud computing. Crypt. Secur. arXiv: 1404.6849 (2014) 10. Peng, L., Song, G., et al.: Privacy—preserving access to big data in the cloud. IEEE Cloud Comput. 3(5), 34–42 (2016). https://doi.org/10.1109/MCC.2016.107 11. Mohammed, M.B.: Security and privacy challenges in mobile cloud computing: survey and way ahead. J. Netw. Comput. Appl. 84, 38–54 (2017) 12. Shugrue, D.: Fighting application threats with cloud-based WAFs. Netw. Secur. 6, 5–8 (2017) 13. Agoulmine, N., Carvalho, C., et al.: State of the art and challenges of security SLA for Cloud Computing. Comput. Electr. Eng. 141–152 (2017) 14. Castiglione, A., Narducci, F., et al.: Biometric in the cloud: challenges and research opportunities. IEEE Cloud Comput. 4(4), 12–17 (2017) 15. Varadharajan, V., et al.: Securing services in networked cloud infrastructures. IEEE Trans Cloud Comput. 6(4), 1149–1163 (2016). https://doi.org/10.1109/TCC.2016.2570752 16. Benzeki, K.: A secure cloud computing architecture using homomorphic encryption. IJACSA 7(2), 293–298 (2016) 17. Naskos, A.: Online analysis of security risks in elastic cloud applications. IEEE Cloud Comput. 3(5). https://doi.org/10.1109/mcc.2016.108, 26–33 (2016) 18. Carlson, F.: Security analysis of cloud computing. Cryptography and Security. Article no 261952355 (2014) 19. Rong, C.: Beyond lightning: a survey on security challenges in cloud computing. Comput. Electr. Eng. 39(1), 47–54 (2013) 20. Zang, Y.: Cryptographic public verification of data integrity for cloud storage systems. IEEE Cloud Comput. https://doi.org/10.1109/mcc.2016.94 (2016) 21. Jiawei, Y., Schucheng, Y.: Privacy preserving back-propagation neural network learning made practical with cloud computing. IEEE Trans. Parallel Distrib. Syst. 25(1), 212–221 (2014) 22. Aguiar, Z.: An overview of issues & recent developments in cloud computing & storage security. In: Part 1 Paper Presented at the High-Performance Cloud Auditing and Application, pp. 1–31. Springer, NewYork, London (2013) 23. Balamurugan, B.: Is cloud secure. IJCSE 4(10), 126–129. ISSN: 2347-2693 (E) (2016) 24. Bohli, J., Nils, M.: Security and privacy-enhancing multicloud architectures. IEEE Trans. Dependable Secure Comput. 10(4), 212–224 (2013) 25. Pearson, S.: Privacy, security and trust in cloud computing. Priv. Secur. Cloud Comput. 3–42 (2013) 26. Nepal, S.: TruXy: trusted storage cloud for scientific workflows. IEEE Trans. Cloud Comput. 5(3), 428–442. https://doi.org/10.1109/tcc.2015.2489638 (2016) 27. Ryan, M.D.: Cloud computing security: the scientific challenge and a survey of solutions. J. Syst. Softw. 86, 2263–2268 (2013) 28. Teeba, M.: Homomorphic encryption applied to cloud computing security. In: Paper Presented at the World Congress of Engineering, vol. 1, pp. 112–118 29. Shiwali.: Providing arithmetic operations on RSA using homomorphic technique. IJCSE 4(9), 43–47. ISSN: 2347-2693 (E) (2016) 30. Parmar, P., Shraddha, P.B., et al.: Survey of various homomorphic encryption algorithms and schemes. Int. J. Comput. Appl. 91(8), 26–32 31. Rivest, R.L., Adleman, L., Dertouzos, M.L.: On data banks and privacy homomorphisms. Found. Secur. Comput. 11, 169–180 (1978)

92

B. Seth et al.

32. Rani, B.: A novice’s perception of partial homomorphic encryption schemes. Indian J. Sci. Technol. 9(37). https://doi.org/10.17485/ijst/2016/v9i37/87977 (2016) 33. ElGamal, T.: A public key cryptosystem and a signature scheme based on discrete logarithms. In: Advances in Cryptology, pp. 10–18. Springer (1985) 34. Pascal, P.: Public key cryptosystems based on composite degree residuosity classes. Adv. Cryptol. EUROCRYPT’99 1592, 223–238 (1999) 35. Kumar, R., Verma, H.K., Dhir, R.: Analysis and design of protocol for enhanced threshold proxy signature scheme based on RSA for known signers. Wirel. Pers. Commun. Int. J. 80(3), 1281–1345 (2015). ISSN 0929-6212 (Print) 1572-834X (Online) 36. Jost, C.: Encryption performance improvements of the Paillier cryptosystem. Int. Assoc. Cryptol. Res. Cryptol. ePrint Arch. (2015) 37. Xiao, Z., Yang, X.: Security and privacy in cloud computing. IEEE Commun. Surv. Tutor. 15, 843–859 (2013) 38. Kumar, R.: Cryptanalysis of protocol for enhanced threshold proxy signature scheme based on elliptic curve cryptography for known signers. Knowl. Comput. Appl. https://doi.org/10.1007/ 978. ISBN 978–981-10-6679-5. Springer Nature Singapore Pte Ltd. (2018)

A Gamification Framework for Redesigning the Learning Environment Rishipal, Sweta Saraff and Raman Kumar

Abstract In the modern education system, one of the major concerns today is maintaining students’ interest in the school curriculum. With the growing popularity of gamification in various fields, it can serve as a useful pedagogical tool for the educators to make the learning environment more stimulating and immersive. There is a need for a succinct and coherent framework of gamification. Though gamification cannot be a panacea for all maladies, our focus is on crafting a research based and carefully designed model, so that we can utilize it to its maximum potential.

Play is innate to not only humans but also animals. In any “Play”, the primary reinforcement is the fun element or enjoyment. There is a freedom of choice, a sense of autonomy as game selection depends on the player’s interest, personality and level of difficulty. This empowerment is one of the basic premises on which the concept of “Play” lies on. “Play is not an optional leisure activity, but a biological imperative that supports our cognitive and emotional well-being, occupying an important role in our development as humans” [1]. Often players are so engrossed in games that they enjoy the challenges, simply to achieve the desired targets and collect badges. Sometimes the game is so immersive that they become emotionally charged. The players sometime develop a feeling of affiliation and sentimental attachment to the co-gaming peer group. A game is more structured than a play with clearly identified rules and goals. Players engage in conflicts or challenges, even in a single user game they try to boost their last scores. Learning is usually not considered fun or entertaining, therefore “It is much more challenging to keep the students motivated to engage in studies,” [2]. It is difficult to keep learners motivated in the process of acquiring a new skill or Rishipal (B) · S. Saraff AIBAS Amity University, Gurgaon, Haryana, India e-mail: [email protected] R. Kumar Department of Computer Science and Engineering, I K Gujral Punjab Technical University, Kapurthala, Punjab, India e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_6

93

94

Rishipal et al.

securing knowledge [3] as if they are playing a game. Combining the best of both learning and gaming may draw the attention of passive learner also. Gamification is using game mechanics and elements in a non game context. Gamification, defined as “the use of video game elements in non-gaming systems to improve user experience and user engagement” [4], has been widely accepted in training purposes. It has become so popular, that an author goes so far as to say that gamification is “Coming soon to your bank, your gym, your job, your government and your gynaecologist” [5]. Kumar et al. (2015) proposed a proxy signer for recovering the shares from another proxy server shares, without capturing the information about shares [6, 7] According to Zichermann and Cunningham [8] “Gamification is the process of using game thinking and game mechanics to solve problems and engage users”. Game mechanics make learning and instruction more enjoyable and interesting [9]. Kapp [10] states the utility of a gamified environment in providing immediate feedback to students about their current progress as well as acknowledging their achievements. Also, it has a great capacity for increasing engagement and motivation of students in the classroom [11]. The elements of fantasy, curiosity and challenge attract the player’s cognizance, where users remain absorbed in the game [12]. The term “Gamification” caught attention with the publishing of books about the use of game mechanics in different areas. In “Reality is Broken”, by Jane McGonigal [13], the author argues that “The use of games does not need to be focused only for entertainment and that skills developed during the game are useful to solve real-life issues”.

1 The Evolution of Gamification in Education Nick Pelling coined the word “Gamification” in 2002 [14]. He was a game developer from Britain, who wanted to use entertainment programmes as promotional strategies. Later Rajat Paharia [15] in October 2007, with the support of Adobe Systems [16] started using game mechanics as a marketing tool. Khan Academy [17] gamified education with the purpose of maintaining students’ interest and Codeacademy [18] also used gamification brilliantly to engage students in learning coding. Meanwhile the first Gamification Summit was held in San Francisco in 2010. This event was attended by McGonigal and Zichermann. As already discussed, the word “gamification” became popular after Deterding (2011) conceptualized its usage in non gaming contexts [19, 20]. Several empirical studies were being conducted to determine its effectiveness in various contexts [21]. There were many controversies that sparked, whether this approach would stand the test of time and sustain interest of the users? Its potential in the field of education was being widely contemplated by researchers, educationists, theorists, etc. Motivation is being considered as the key component of academic success which impacts students’ learning behaviour. The cognitive–emotional perspective not only gave a rational view point but also added the element of dynamism to it.

A Gamification Framework for Redesigning the Learning …

95

2 Does Gamification Differ from Serious Games? Serious games incorporate elements of game in the context of gaming only [4] where as gamification is the use of use of game mechanics in contexts other than games. Kapp [10] encouraged application of storytelling, problem solving, interaction, characterization and commitment as the basic premise on which the concept of gamification can be build. He was an advocate of using creativity and reasoning abilities to motivate and engage users. He encouraged use of badges, points and rewards to build a strong connect with students but also explicitly stated that game features should not be limited to them only. Although he accepted that creating an engaging gamified learning system is both rigorous and demanding but strongly discouraged undervaluation of curriculum and any sort of compromise with learning material. Also he never believed that it is “perfect for every learning situation” and considered a methodological and insightful approach to be powerful and long withstanding from an educational point of view.

3 Theoretical Background Self-Determination Theory gives an insight into the conceptual background of gamification [22]. “SDT is an approach to human motivation and personality that uses traditional empirical methods while employing an organismic meta-theory that highlights the importance of humans’ evolved inner resources for personality development and behavioral self-regulation” [23, 24]. Thus, it investigates people’s psychological tendencies as the basis for their self-motivation and personality development, as well as the environment that promotes these positive processes. Autonomy, relatedness and competency are the core elements required for optimal functioning, constructive development and self- integration. While explaining the SDT theory, Deci and Ryan differentiates between three main types of motivational states, namely ‘intrinsically motivated, extrinsically motivated and amotivated states’ [25] (Fig. 1). When learners are unmotivated, they are apathetic towards most of educational activity. Autonomous motivation consists of both intrinsic and internalized extrinsic motivational states [26]. An intrinsically motivated student has a passion for learning, is incessantly curious and enjoys the process more than the outcome. Whereas, some students foresee the utility of a particular course and the prospective material gains. Psychological wellness, use of meta-cognition (meticulous planning and managing time), will power and perseverance, in-depth learning and seeking high grades are usually associated with autonomous motivation [27].

96

Rishipal et al.

Types of Motivation Intrinsic

Extrinsic Motivation

Motivation

Amotivation

Intrinsic

Identified

Introjected

External

No

Regulations

Regulations

Regulations

Regulations

Motivation

Autonomous Motivations Controlled Motivations Fig. 1 The different types of motivation (based on [25], p. 16)

4 Concept of Flow Mihaly Csikszentmihalyi was the first to study the concept of flow and got amazed by the fact that a painter gets so immersed in the painting process that he even forgets his basic needs [28, 29]. Based on their studies, Csíkszentmihályi and Nakamura [30] discovered that the concept of “being in flow” consists of the combination of the six following experiences: • • • • • •

intense and focused concentration on the present moment merging of action and awareness loss of reflective self-consciousness sense of personal control or agency over the situation or activity distortion of temporal experience, one’s subjective experience of time is altered experience of the activity as intrinsically rewarding, also referred to as autotelic experience.

5 Gameflow Sweetser and Wyeth [31] suggested a new model for the concept of flow—“Gameflow consists of eight elements—concentration, challenge, skills, control, clear goals, feedback, immersion, and social interaction”. Chen [32] lays out the basic nuances while designing a game. The game must be engrossing and the right balance between difficulty level and users’ sentiments must be maintained. The player loses interest if the game is too easy or too complicated. Thus in order to keep the player in the “Flow”, the game must be exciting. The same construct can be applied to learning as well (Fig. 2).

A Gamification Framework for Redesigning the Learning …

Action opportunities (Challenges)

97

Anxiety Flow Boredom

Action capabilities (skills) Fig. 2 The flow channel is a balance between challenge and skill. Anxiety arises when challenges cannot be met with one’s skills. If challenges are inadequate for one’s skills, boredom accrues. Adapted from “The concept of flow”, by Csikszentmihalyi and Nakamura [30], Handbook of positive psychology, p. 94

6 From Socio-emotional Perspective Most of the research on gamification has concentrated on both affective and social components but the most predominant attention has been given to motivational aspect. According to Zichermann and Cunningham [8], use of social elements is noteworthy for people who are active on social platforms as certain features of gamification are also shared by social games for example, leaderboards, levels, virtual currency, etc. These platforms encourage knowledge- sharing, interpersonal relationships and perspective taking. Considering the inherent challenge in developing the social elements in gamification, Simoes et al. [11], used an existing social learning platform for K-6 which can be efficiently “integrated with new gamification trends”. Team work and cooperation are the most alluring features of a social game, it imbibes relatedness and heightened emotional reactions to both achievements and failures if the game has succeeded in captivating users’ interest. Human behavior is explained as “complex and blurred interactions along multiple effective and cognitive dimensions”, Mullins [33]. Cognition refers to the mental activities pertaining to sensation, perception, attention, memory, language, movement, thinking, reasoning, problem solving and decision making. It has been widely accepted that there is an underpinning of emotions and subjective feelings interacting with these cognitive fields affecting the perception and thinking capacities to a larger extent.

98

Rishipal et al.

Skinner’s theory of operant conditioning and Bandura’s social learning theory explain role of both positive and negative emotion in learning. Often fear of failure and incompetencies stimulate negative emotions and associated physiological reactions such as heightened vigilance, perspiration, increased pulse and pressure, etc. These events create episodic memories in amygdala and hippocampus which further results in pathologies of evaluation anxiety. Whereas happiness in small achievements and confidence of getting a second trial or chance inculcate positive self concept and high self esteem in learners. These events stimulate a chain of positive reactions and joy of learning.

7 Gamification of the Learning Environment A well designed learning environment is characterized by three core features i.e. autonomy, belongingness and competency. “Autonomy” is the key feature which makes the exercise of gamification relevant and fruitful. It is synonymous with empowerment and there is an inherent freedom of choice. The students feel that the mentors are considerate towards their preferences. The type of choices given can be manifold ex- players can select teams, characters/avatars, topics or projects, date of submission, etc. This structure not only fosters personalized learning but also provides sufficient opportunities for the student to. ruminate over his choices [34]. No one plays the game which doesn’t catch their imagination. Learners feel that they have control over what they want or like to learn. It is more about giving a meaningful and guided opportunity. The next principle is “Belongingness”, and in a good game environment, the relationships which the members share make them tied up to the game. They develop camaraderie, build a fraternity based on shared interest, sometimes there may be an emotional attachment to the game. Cooperation and collaboration are being taught implicitly, when a player’s progress depends upon other team members. Another important feature is the type of feedback they receive; sometimes students do not enjoy leaderboards or social graphs. Some of them may want to work at a relatively slower pace. Game leaderboards sometimes work against a student’s development, as it creates a sense of isolation, where each student competes against the other. The third principle is feeling of competence, if players feel that they have the liberty of choosing among topics and level of challenge, they are encouraged to accept the challenge and work hard towards accomplishing it. As the course levels up, the students are ready to face new challenges. The course designer must remember to maintain a flow wherein the levels must not be extremely difficult and neither very easy. There must be a balance so that the students are able to achieve the desired goals with adequate scaffolding. This is also very much related to the idea from Vygotsky and Social Learning Theory of the Zone of Proximal Development.

A Gamification Framework for Redesigning the Learning …

99

Autonomy, sense of belongingness and feeling of competence reinforce positive learning environment where learners perform better. Here we must also consider productive failure, encouraging exploration and supporting perseverance. Games support and reinforce retrying, there is a freedom to fail and scope to win after targeted practice. Let’s see it this way—if a character dies in a game after falling or not being able to overcome obstacles in a particular level, the player always has a chance to improve his techniques next time. Games do a great job of supporting failure, by not punishing you much if you don’t succeed at something. The cost of losing out is not magnanimous compared to real time school tests. Failure in school is such a huge set back that it may create the possibility of students trying easy courses to obtain higher grades. They develop unproductive habits like cheating, trying to look for shortcuts, things that are not productive to learning environments.

WELL DEFINED LEARNING GOALS

FEEDBACK

FREEDOM TO FAIL

CHOICE

TRANSPERANCY

KEY FEATURES OF GAMIFICATION Use of game mechanics and game dynamics support game thinking by allowing students to reason, contemplate, generate various options, analyse and thereafter synthesize. Game thinking develops advance thinking mechanisms in young learners as they learn how to manipulate their environment with the available resources. Critical thinking involves knowledge, comprehension, application, analysis, synthesis and evaluation [35]. A player never chooses a game, of which he or she has inadequate

100

Rishipal et al.

knowledge, does not understand the rules or feels unrelated. A game can be a puzzle, mission, arcade, adventure or action filled (Google play store) but it must be entertaining, fun filled with adequate levels of difficulty. The problem solving capacities of game based learning enhances students’ performance and helps in stimulating various cognitive operations like working memory and focused attention. Two most important features of problem solving are that it is a conscious activity and goal directed. Games provide a well defined platform for goal setting and incentivize winning by giving ‘xp points,’ trophies, badges, virtual goods/gifts, etc. There is an opportunity for learners to understand their strengths and weaknesses, to improvise and come back again in the same game, same level to raise their score. The key advantage is the ability to leave the game at your own will or play hard to become a leader and develop game status. This environment is in many ways similar to learning environment except there is more freedom and autonomy in games. Today’s children perceive curriculum to be boring, rigid and monotonous, it takes away the joy of learning. The game competency involves knowledge, skill, ability and willingness, like any other professional competency. Rigorous practice, team work and sustained efforts are required in maintaining your name in a popular game’s leaderboards. There are negative aspects to it also. Competitions like ‘Boss battles’ or ‘Team challenges’ can contribute to stress and disrupt learning. The players are stuck to the game, incessantly playing same levels either to defeat the opponent or to achieve the highest score. This situation creates heightened emotionality and affects the basic rationality of enjoyment from playing a game. Some players may perceive humiliation on losing and become judgemental about their capacities or they may waste their time by playing a certain level for long. Game elements must be judiciously selected for gamifiying a particular course so as to avoid the fallacies as much as possible. The idea is to keep the player intrigued and mystified just like an adventurous journey in the path to learning. The element of surprise, level playing field, show of valour and ecstasy of daily achievements maintains the fun score or interest in the game. In the process of gamification, the educator needs to be vigilant and guarded so that the basic agenda of knowledge sharing is not trivialised [36]. The curriculum must be carefully inspected, prior setting of course objectives, maximum time required for completion of each topic and need for adequate scaffolding must be meticulously assessed before planning the layout of ‘gamification framework’. Development of a ‘gamification framework’ requires prior deliberation over the content and the assessments which can be transformed into a game based learning [37]. Starting from the theme, the avatars, points format (scores would be in thousands or hundreds), number of options or the chance allowed, the mystery or the bonus round, peer challenge as well as group activities must be carefully worked out before the game is presented to the students. The teacher must also inform the students how these scores would be converted into exam points, where they can get leverage, importance of participation and scope of learning for all students. It is also the responsibility of the teacher to support students by explaining rules and encouraging them to ask questions, so that their doubts and scepticism are adequately addressed.

A Gamification Framework for Redesigning the Learning … Table 1 Game mechanics and game dynamics (adapted from Bunchball [41]). Game elements

101

Game mechanics

Game dynamics

Points

Reward

Levels

Status

Trophies, badges, achievements

Achievements

Virtual goods

Self-expression

Leaderboards

Competition

Virtual Gifts

Altruism

Gamification Elements Game Mechanic

Game Design

Game Technique

Fig. 3 Gamification elements

The schools need to make students believe that failure is the road to success, they must encourage a ‘win some, learn some’ environment. A good game encourages exploration, it requires players to gather more points and earn badges [38]. Sometimes the players enjoy competition by engaging in leaderboards and trying to win more challenges. Regular practice and reinforcement help students in accomplishing difficult tasks (Table 1). Khaleel et al. [39] employ ‘three gamification elements such as game mechanics (dashboard and progress bar), game design (using a badge as a reward), and game techniques (Leaderboards showing scores or marks)’ as presented in Fig. 3. Gamification of education is redesigning classrooms and assessments to create the school environment more student-friendly. It is more about teaching in a way students want to learn rather than how we prefer to teach. Gamification of the learning environment is not about designing a learning tool or applications; it is about making learning enjoyable and development of skill sets in children which they often miss, when they are looking for shortcuts. Werbach and Hunter [28] is a prominent name in gamification. He organizes game elements into three distinct categories with ‘components at the base, mechanics in the middle and dynamics at top’.

102

Rishipal et al.

Game Dynamics

Game Mechanics

Game Components

Werbach and Hunter [28] proposes following five elements of Game Dynamics: • Constraints are the inherent limitations that will exist in every design process as it impossible to include almost everything in a particular game. • Emotions are the sentiments which keep the players attached to the game. • Narrative is the story line that keeps the player intrigued. • Progression defines the player’s understanding and level of engagement. • Relationships are the most significant element. Players develop a bond with other players or even the virtual characters in the game. They eventually identify themselves with those “Avatars”. Game Mechanics form the second group of elements which are the basic driving force. “The ten game mechanics are challenges, chance, competition, cooperation, feedback, resource acquisition, rewards, transactions, turn and win states” Werbach and Hunter [28]. The largest group of elements are the Game Components. These are more concrete in nature and form the pillars of the game design. There are innumerable Game Components; some of the popular ones are as follows: Avatars, Badges, Points, Tokens, Combat, Boss Battles, Levels, Leaderboards, Quests, Teams, Virtual goods/gifts, etc. A game designer (class teacher, business manager, etc.) can follow the following steps to make their classroom interesting: 1. 2. 3. 4.

Clear definition of learning objective. Outline or sketch out specific target behaviours/assessment goals. Define players/learners Create interesting activities meticulously keeping track of the beginning, the end and intermittent loops. 5. Incorporate FUN

A Gamification Framework for Redesigning the Learning …

103

6. Make judicious use of Game Mechanics and Components. Too much of anything can make the game boring. Here is an example of gamifying a class VIII/IX classroom. Below, you’ll see there are three levels of objectives.

Subjects

Basic

Intermediate

Proficient

English/hindi

Class attendance

Creative essays

Blogging

Class work

Journal diary

Project on reflective thinking on a book

Participation in reading text

Grammar activity

Debate peer challenge

Class attendance

Home work

Teacher challenge questions

Class work

Class quiz

Peer challenge

Queries/attempt to answer

Written assignments

Team challenge

Class attendance

Home work

Group project

Class work

Class quiz

Model development/basic

Participation in lab activity

Project work

Research/critical analysis of any current topic

Class attendance

Home work

topic/reporter’s diary

Class work

Class quiz

Political debate

Participation in discussion

Project work

Battle of the empires quiz

Maths

Science

Social science

The course instructor can use software applications like Edmodo, Google Classroom or Quill to provide instant feedback to students about their progress. Assessments are considered as levels and students are given points or scores instead of marks or grades. Teachers can become creative by giving badges or letting students to choose a specific character or avatar. They can design a customized leader-board for the classroom, keeping it optional for students to participate. Gamification encourages participation over winning [40]. A well-designed game just like any other application must be dynamic in nature so as to maintain the curiosity of the player. The socio-cultural element inspires participation of players by providing a platform which encourages community building, inculcates leadership skills, creates a level playing field for every learner and nurtures peer relationships. Here instructors also get the liberty to think beyond the boundaries of the curriculum, be innovative with class work and assessment by working closely and in tandem with the students. The basic premise of gamification rests on making school a game of learning and not an endless maze.

104

Rishipal et al.

References 1. Galarneau, L., Zibit, M.: Online games for 21st century skills. In: Gibson, D., Aldrich, C., Prensky, M. (eds.) Games and Simulations in Online Learning: Research and Development Frameworks. Information Science Publishing, USA (2007) 2. Prensky, M.: The motivation of gameplay: the real twenty-first century learning revolution. On Horiz. 10(1), 5–11 (2002) 3. Paras, B., Bizzocchi, J.: Game, motivation and effective learning: an integrated model for educational game design. In: Changing Views: Worlds in Play, Conference of the Digital Games Research Association, 16–20 June. Vancouver, BC (2005) 4. Deterding, S., Sicart, M., Nacke, L., O’Hara, K., Dixon, D.: Gamification: toward a definition. In: Proceedings of the CHI 2011 Gamification Workshop, Vancouver, British Columbia, Canada (2011) 5. Robertson, M.: Can’t play, won’t play. Retrieved from http://www.hideandseek.net/2010/10/ 06/cant-play-wont-play/ (2010) 6. Kumar, R., Verma, H.K., Dhir, R.: Analysis and design of protocol for enhanced threshold proxy signature scheme based on RSA for known signers. Wirel. Personal Commun. Int. J. 80(3), 1281–1345 (2015). ISSN 0929-6212 (Print) 1572-834X (Online) 7. Kumar, R.: Cryptanalysis of protocol for enhanced threshold proxy signature scheme based on elliptic curve cryptography for known signers. Knowl. Comput. Appl. https://doi.org/10.1007/ 978 (ISBN 978-981-10-6679-5). Springer Nature Singapore Pte Ltd. (2018) 8. Zichermann, G., Cunningham, C.: Gamification by Design: Implementing Game Mechanics in Web and Mobile Apps. O’Reilly Media, Sebastopol, CA (2011) 9. Zichermann, G., Linder, J.: Game-Based Marketing: Inspire Customer Loyalty Through Rewards, Challenges, and Contests. Wiley, Hoboken, NJ (2010) 10. Kapp, K.M.: The Gamification of Learning and Instruction: Game-Based Methods and Strategies for Training and Education. Pfeiffer, San Francisco, CA (2012) 11. Simões, J., Redondo, R.D., Vilas, A.F.: A social gamification framework for a K-6 learning platform. Comput. Hum. Behav. (2012) 12. Kirriemuir, J., McMarlane, A.: Literature Review in Games and Learning Literature Review in Games and Learning (2004) 13. McGonigal, J.: Reality is broken: Why games make us better and how they can change the world. Penguin, London (2011) 14. Pelling, N.: The (short) prehistory of gamification, Retrieved from http://nanodome.wordpress. com/2011/08/09/the-shortprehistory-of-gamification/ (2011) 15. Bunchball (2012) Enterprise Gamification: Engaging & Motivating the Gen Y Workforce. Retrieved from http://www.bunchball.com/resources/enterprisegamification, 10 January 2015 16. AdobeSystems. http://www.adobe.com/ (2012) 17. KhanAcademy.: KhanAcademy. http://www.khanacademy.org/ (2012) 18. Codeacademy. http://www.codecademy.com/ (2012) 19. Deterding, S.: Situated motivational affordances of game elements: a conceptual model. In: Gamification: Using Game Design Elements in Non-Gaming Contexts, a Workshop at Chi (2011) 20. Deci, E., & Ryan, R. (2012). Overview of Self-Determination Theory. In: The Oxford Handbook of Human Motivation, p. 85 21. Seaborn, K., Fels, D.I.: Gamification in theory and action: a survey. Int. J. Hum. Comput. Stud. 74, 14e31 (2015) 22. Csikszentmihalyi, M.: Finding Flow: The Psychology of Engagement with Everyday Life. Basic Books (1998) 23. Ryan, R.M., Deci, E.L.: Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. Am. Psychol. 55(1), 68–78 (2000) 24. Ryan, R.M., Kuhl, J., Deci, E.L.: Nature and autonomy: organizational view of social and neurobiological aspects of selfregulation in behavior and development. Dev. Psychopathol. 9, 701–728 (1997)

A Gamification Framework for Redesigning the Learning …

105

25. Deci, E., Ryan, R.: Handbook of Self-Determination Research. University Rochester Press (2004) 26. Deci, E., Koestner, R., Ryan, R.: A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation. Psychol. Bull. 125(6), 627 (1999) 27. Deci, E., Koestner, R., Ryan, R.: Extrinsic rewards and intrinsic motivation in education: Reconsidered once again. Rev. Educ. Res. 71(1), 1–27 (2001) 28. Werbach, K., Hunter, D.: For the Win: How Game Thinking Can Revolutionize Your Business. Wharton Digital Press, Philadelphia (2012) 29. Csikszentmihalyi, M.: Flow: The Psychology of Optimal Experience. Harper Perennial, New York (1990) 30. Csikszentmihalyi, M., Nakamura, J.: The Concept of Flow, the Handbook of Positive Psychology. Oxford University Press pp. 89–92, ISBN 9780195135336 (2002) 31. Sweetser, P., Wyeth, P.: GameFlow: a model for evaluating player enjoyment in games. Comput. Entertain. 3(3), 3 (2005). https://doi.org/10.1145/1077246.1077253 32. Chen, J.: Flow in games (and everything else). Commun. ACM 50(4), 31–34 (2007) 33. Mullins, J.K.: Beyond Enjoyment: A Cognitive-Emotional Perspective of Gamification, pp. 1237–1246 (2018) 34. Reeve, J., Jang, H.: What teachers say and do to support students’ autonomy during a learning activity. J. Educ. Psychol. 98, 209–218 (2006) 35. Adesoji, F.A.: Bloom taxonomy of educational objectives and the modification of cognitive levels. Adv. Soc. Sci. Res. J. 5(5) 36. Pessoa, L.: The cognitive-emotional brain: from interactions to integration. MIT press, Cambridge (2013) 37. Blohm, I., Leimeister, J.M.: Design of IT-based enhancing services for motivational support and behavioral change. Bus. Inf. Syst. Eng. 5(4), 275–278 (2013) 38. Hamari, J.: Do badges increase user activity? A field experiment on the effects of gamification. Comput. Hum. Behav. 71, 469–478 (2017) 39. Khaleel, F.L., Ashaari, N.S., Wook, T.M., Ismail, A.: Gamification elements for learning applications. Int. J. Adv. Sci. Eng. Inf. Technol. 6(6), 868. https://doi.org/10.18517/ijaseit.6.6.1379 (2016) 40. da Rocha Seixas, L., Gomes, A.S., de Melo Filho, I.J.: Effectiveness of gamification in the engagement of students. Comput. Hum. Behav. 58, 48e63. https://doi.org/10.1016/j.chb.2015. 11.021 (2016) 41. Bunchball Inc.: Gamification 101: an Introduction to the Use of Game Dynamics to Influence Behavior. Available on http://www.bunchball.com/gamification/gamification101.pdf (retrieved on 27.07.2011) (2010)

A Variant of Secret Sharing Protected with Poly-1305 Shakti Arora, Surjeet Dalal and Raman Kumar

Abstract We are working in the era of cloud computing, where all of the required resources are available online at pay-par-basis. It made all the IT industry easily accessible to all types of users. It provides the services in software, hardware and in storage terms. We are dealing the model of IAAS which provides on demand secured storage services. A number of researchers has designed and proposed a number of techniques and algorithms for assurance of storage services provided by cloud service providers. Our paper presents a modified approach of integrity verification in multiparty communication in decentralized cloud computing environment. We enhanced the basic model of AES with AES Poly Library 1305 and also redesigned the variant of secret sharing scheme for handling a secured group communication. Our factors for evaluation are the hardness and randomness of key i.e. entropy of the proposed technique and other measurable units which gives the efficiency of communication overhead and security. Keywords Entropy · CPU cycles · Poly-1305 · Encryption · Decryption

1 Introduction A new revolution and advancements in field of hardware, software, virtual machines, and middleware or in can say revolutionary advancements in IT technology has led to an emergence of wide global distributed platform. Now a days cloud computing S. Arora (B) · S. Dalal SRM University, Sonepat, Haryana, India e-mail: [email protected] S. Dalal e-mail: [email protected] R. Kumar Department of Computer Science and Engineering, I K Gujral Punjab Technical University, Kapurthala, Punjab, India e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_7

107

108

S. Arora et al.

proves a storage. As a service which is accessible from anywhere via internet connection without imposing and spending a higher cost in set up. In the same way lot of investments are done in setup an infrastructure and purchasing a licensed software is drastically decreased and usage of IT technologies is increased but the overall cost of using the technology is decreased, which makes it easily accessible to every person. According to survey done it has been find out that 63% of financial services, 62% of manufacturing service, 59% of healthcare service and 41% of transportation industries are getting advantage of cloud computing applications. According to rackspace, the pay as you go service model saves 58% of cost. As a result 70% of global companies projected to store their important and crucial data on the cloud latest by 2018 [1].

2 Literature Review Jianhang et al. (2010) proposed an integrity verification scheme on the concept of RSA security. The biggest advantage of this new technique achieved that it does not store the data on client machine and does not load the entire data from the server to local machine for data integrity authentication. It involved a third party called PKG cryptography to check the data integrity at server side. PKG plays the role of trusted third party to which client delegates the authority of check integrity. It was believed that client is secured and PKG is trusted [2]. Wang et al. (2011) proposed a approach for dynamic data operations and public audit ability. Other designed approaches was depending on the assurance of third party verification, what if the third party is compromised, This proposed approach focus on the public audit ability of data in dynamic operations of data [3]. Sravan Kumar R. et al. (2011) developed an integrity verification technique which generated the proof of data integrity in the cloud which can be used by the customers to believe and get assured about the integrity of data residing in the cloud. Integrity proof is incorporated in the service level agreement i.e. SLA. Main advantage achieved with this technique is that client storage is at minimal use. Data could not be altered or modified without the permission of data owner, because of frequent checks applied at different interval of time, these integrity checks allowed the data owner to do proper verifications efficiently [4]. Zhuo Hao et al. (2011) proposed a technique without involvement of third party auditor; it started with the live protocols that are working with the TPA. Currently, the mechanisms hold a public auditing and verifiability by involving a third party auditing. This paper holds public verifiability without the help of third party. No sensitive information is leaked to third party [5]. M. Venkatesh et al. (2012) proposed an improvised RSA storage security, which used the concept of public auditing of the remote data by upgrading the RSA based signature generation. The purpose of RSASS (RSA based signature security) is to provide data storage correctness, and identifies misbehaved server. Dynamic operations on the data is supported by this technique so computation complexity is reduced.

A Variant of Secret Sharing Protected with Poly-1305

109

Public verifiability is allowed means any client can verify the data present on the cloud, no personal information is retained by cloud server. RSASS is a method chosen for auditing the data place on the cloud server. TPA ensures and check the data file as well as keeps monitoring on the data available on the cloud server for the assurance of clients [6]. Louis Goubin et al. (2011) proposed an original masking scheme based on Shamir secret sharing as an alternative to Boolean masking. This scheme was implemented with AES in multi party computations. Implementation of the proposed scheme provides the security level of 3rd level masking together with a better efficiency [7]. Zhifang Zhang et al. (2009) proposed a change in threshold value of secret sharing scheme. Threshold changeable schemes turn out to be insecure under the collusion attack of players holding initial shares, he constructed a computationally secure scheme under the enhanced model with much shorter shares and broadcast message than the perfect scheme [8]. S. Jaya Nirmala et al. (2012) did comparative analysis of two different secret sharing schemes and applied the best secret sharing scheme with robin dispersal information algorithm in private cloud set up using open stock, he concluded that information dispersal of data is the more optimal approach for data outsourcing. Abbott et al. (2007) performed attack simulation and traffic analysis on browser level, proposed technique was working on TOR exit system to analysis the traffic [9]. Aberer and Despotovic (2001) presented a reputation based trust management at data level and semantic level. An agent management system was designed to calculate the trust [10]. Kumar et al. (2015) proposed a proxy signer for recovering the shares from another proxy server shares, without capturing the information about shares [11, 12]. Armbrust et al. (2009) discussed the concept of cloud computing with real life applications, a comparative analysis is designed with traditional system and changes of technology in virtualization [13]. Acharya et. al (2000) designed a suite of untrusted application which can rectify the behavior of applications according to stored classes and objects [14]. Akug et al. (2008) established techniques of security and safety of documents, monitoring of inline java virtual machine, and embed a number of security API to check the security parameters [15]. Kumar et al. [2007–2016] given method for RSA and Lagrange Coefficient [11]. Kumar et al. [2017–2018] given procedure for ECC and Lagrange Coefficient [12].

3 Research Gap A number of research papers and articles are published and I have mentioned the important papers in the literature and analyzed few challenges or the issues that need to be focused. Cryptography is wide area, and in daily routine a number of techniques

110

S. Arora et al.

have been proposed to cover the latest issues of security. Usage of cloud computing and security of data at distant location which is not known to clients and data users makes the information more at risk levels [7]. Following Gaps in the literature has been identified to take the research one step forward (1) AES is used for strong encryption of data and widely adapted in the cloud environment, we used the AES technique to be merged with SSS is resolving side channel attack (2) Computation complexity of the above implemented algorithm is very high (3) In classic approach of Shamir Secret Sharing integrity check at multiparty is not implemented properly. (4) Short length passwords are the challenges for encryption applications. And it’s difficult to remember the long password. (5) Trust factor calculation is significantly dependent on reputation of cloud service provider, some factors need to be numerically calculated to achieve and increase the trust level.

4 Problem Statements Trust is an entity which depends on multi valued attributes like honesty, reliability, truthfulness. Security, QOS, time-lines, dependability and return on Investment. Cloud Integrity includes Privacy, security, consistency and accuracy of data. The current widely used clouds like Amazon, IBM, Microsoft have provided a universal trust level in centralized way, still the majority of security issues arises due to running and keeping large sized sensitive data on the cloud. If any of the node in communication system is compromised then, the whole computation is compromised or threat in data integrity. For e.g. in order to satisfy customer needs from anywhere the information posted by the customer is not maintained by single site or computer, rather maintained in a number of trusted nodes. If only one gets failed then other one can provide the services [16]. Few of the upgraded techniques and algorithms are working, with good efficiency and provide consistency of data but it had also added the computation and communication overhead in the system which makes it unrealistic to use them in real time applications We have proposed and focused on the working of decentralized cloud computing, where multi party communication happen, and instead of protecting the centralized server we are trying to maintain the integrity of data at lower level decentralized nodes/workstations/Server to improve security without impairing efficiency Cryptographic algorithm implemented with hardware or physical devices are prone to side channel resistant (SCR). The most common technique for block cipher is to use masking which produces random variables and apply protection.

A Variant of Secret Sharing Protected with Poly-1305

111

To propose a masking scheme based on Shamir’s Secret Sharing scheme as an alternative to Boolean masking. AES with secret sharing scheme (AES-SSS) without the use of a cryptographic key is implemented. We applied variant on both the techniques in combination to achieve the desired goal [7]. • To achieve the assurance of integrity of sensitive data on different servers of decentralized cloud computing. • To handle communication overhead in cloud computing environment. • Performance evaluation of proposed scheme with different security parameters.

5 Objectives In cloud computing environment, huge volume of data is stored and travel around the internet, from where we are storing and retrieving the data is purely hidden from the client and user. There is no single mechanism or technology that can make the whole system attack proof. In order to protect the business enterprises and government sensitive data a number of techniques are available like SLA, RSA, DES, etc. to secure the data [17]. These security measures of technologies are integrated into cloud computing architecture. We know the development from the scratch is quiet difficult. So we are going to use re-engineering methodology. Our main objectives of the proposed system are set as: • To design an architecture for cloud computing environment and integrate an improved integrity algorithm for integrity verification • Review the different algorithms/techniques of cryptography and come up with a new way to maintain the integrity of data and reduce the computation time taken by algorithms. • To handle the computation integrity metrics efficiently.

6 Existing Methodology 6.1 AES with Classical Shamir Secret Sharing AES is block cipher that iterates ten, twelve and fourteen cycles of repetitions of transformation rounds, in each of the transformation round it involves four different stages of operations • • • •

ADD Round Key Shift Rows keys Mix Columns Sub Bytes

112

S. Arora et al.

Above four operations are performed in each iteration that provides the enhanced security level. It performs the encoding of secret Key Kb to be an arbitrary hexadecimal string of length l, result a random hexadecimal number H1 ; H2 ; Hn , whose length is equal to the size of secret key Kb, which is of length 1. The last participant receives the result by performing the following XOR function. For recovering the secret it requires partial information regarding the shares. Variation in XOR function requires that all of the shares be pooled together to recover the secret Kb. A masked secret and its d + 1 shares are given as input to the algorithm. At the beginning of the operation/computation, the secret is divided into d + 1 share, with respect to secret sharing scheme.

6.2 Existing Algorithm Step 1: Plaintext_Hexsecretsharer (string) Converting secret messages into standard hex shares secret_charset  string share_charset  string.hexdigits [0:16] Step 2: Split_secret (secret_string, threshold, num_shares) secret_int  charset_to_int(secret_string, secret_charset) points  secret_int_to_points (secret_int, threshold, num_shares) shares  [] for point in points shares.append (point_to_share_string (point, share_charset)) return shares Step 3: Masking AddRoundKey Input:- Perform XORing the rth round key kr Output:- d + 1shares (xi ; kr;i )i for every round key kr . Step 4: Recover_secret(shares) points  [] for share in shares: points.append (share_string_to_point (share, share_charset)) secret_int  points_to_secret_int (points) secret_string  int_to_charset (secret_int, secret_charset) return secret_string

6.3 Limitations of Existing Algorithms Classical Shamir secret sharing produce shares which are shorter in length. An attacker could tamper with some of the shares. After restoring the (malicious) secret, you would not be able to know that it has been tampered with as shown in Fig. 1.

A Variant of Secret Sharing Protected with Poly-1305

113

As shown in figure secret message “This is password” is split into 3 shares with threshold 2. Share 2 is tempered, when all the three shares are combined to recover secret message, it recovers the wrong secret message, Shamir secret sharing is secure for confidentiality, this is not the case for integrity.

7 Proposed Algorithm As we found that in decentralized cloud computing environment Shamir secret sharing approach is the efficient approach for sharing the sensitive data with number of users but we find out that there is lot of drawback in classic secret sharing approach. So we proposed some modification in the classic technique and reduced the limitations of the same. AES is widely adopted algorithm for encryption in any of the cloud service provider for example AMAZON and IBM is widely using it for high security reason. AES is affected with brute force attack in 2005 because of flaws in implementation and lack of key management. Here we combined AES and modified secret sharing approach which removes all the flaws of both approaches.

7.1 Targeted Outcomes to Achieve (1) (2) (3) (4)

Dynamic share updating with each secret upload and retrieval Integrity verification and identification at each level of shares collection Usage of weak password and short data length is no limitation Calculate the Entropy of proposed algorithm (compare the result with existing and modified approach) (5) Key generation time in terms of CPU cycles and compare it with the existing approach (6) Encryption time

Fig. 1 Tempered share in Shamir secret sharing

114

S. Arora et al.

(7) Decryption Time (8) Dynamic security of POLY library 1305 MAC with Default AES.

7.2 AES-Poly 1305 with Modified Secret Sharing Scheme 7.2.1

Algorithm Secret-Share-Split

Step 1: Plaintext_Hexsecretsharer (secret_message) Converting secret messages into standard hex shares secret_charset  string share_charset  string.hexdigits [0:16] Step 2: Split_secret (secret_string, threshold, num_shares) secret_int  charset_to_int(secret_string, secret_charset) points  secret_int_to_points (secret_int, threshold, num_shares) shares  [] for point in points shares.append (point_to_share_string (point, share_charset)) return shares Step 3: hkey  (rf, secret_message, Salt, c, dklen) Where rf is a pseudorandom function Secret_message is from which a derived key is created Salt is a series of bits, recognized as a cryptographic salt c is desired number of iterations dklen is the preferred size of the derived key hkey is the produced derived key Step 4: create_keyshares(&key, count, threshold)

7.2.2

Algorithm Secret-Share-Combine

Step 1: Read each line, reading shares into memory Step 2: Decode the lines If line.len () % 2 ! 0 display share is of an incorrect length exit Step 3: Split off the keyshares Extract keyshares from ciphertexts keyshares  with_capacity(decoded_lines.len()); ciphertexts  with_capacity(decoded_lines.len());

A Variant of Secret Sharing Protected with Poly-1305

115

Fig. 2 Modified Shamir secret sharing

Fig. 3 Keyshares

Fig. 4 Tempered shares .txt File

Step 4: Check if none of the shares are corrupted Display Error if the ciphertexts are not all the same exit Step 5: Restore the encryption key key  match combine_keyshares(&keyshares).

7.3 Integrity Verification of Shares with Modified Algorithm Here same secret message is split into 3 shares with threshold 2 and shares are stored in shares .txt file [8].

116

S. Arora et al.

Fig. 5 Integrity verification of shares

As shown in above figure share 2 is tempered (Figs. 2, 3, 4 and 5).

8 Evaluation Parameters 8.1 Entropy The entropy of a document is an index of its information content. The entropy is measured in bits per character. From the information theory point of view, the data in the current window can be viewed as a message source. To calculate the information content one examines the probability distribution of this source. It is assumed here that the individual messages (characters in the document/file) are stochastically independent of each other and are transmitted by the source with a uniform probability.

8.2 Encryption/Decryption Time Encryption time is termed as the time taken to generate cipher text from the given plaintext of an algorithm. Decryption time is the time taken by an algorithm to generate plaintext to from the cipher text.

8.3 Encrypted File Size (B) Size of Cipher text generated.

A Variant of Secret Sharing Protected with Poly-1305

117

Fig. 6 AES entropy

8.4 Key Generation Time Key generation time of AES and AES POLY 1305 has been calculated to find out the time of generating and encrypting the shares. Calculated time is measured in CPU cycles taken by both algorithms

9 Security Analysis Good sources of randomness are essential in cryptography, and entropy is often used to quantify randomness. Low entropy means that your source probably isn’t really random. In particular, if we have a random variable X that takes on values x 1 , …, x n with probabilities p(x 1 ), …, p(x n ) respectively, then the entropy of X is H (X )  −

n 

p(xi ) log p(xi )

i1

This value is maximized when all of the probabilities are the same. If we have 2n different symbols, that maximum value will be n bits of entropy per symbol. That’s the theoretical maximum level of entropy that we can get (Figs. 6 and 7).

10 Conclusions Masking scheme based on Shamir’s Secret Sharing scheme an alternative to Boolean masking is proposed. AES-SSS masking is centered on the signal-to-noise ratio (SNR) generated by the crypto application. Applications involving smart card implementation tend to have a higher SNR, and first-order of SSS masking provided better

118

S. Arora et al.

Fig. 7 AES-Poly-1305 entropy

security than third-order Boolean masking as shown by entropy parameter. We also designed a variant of Shamir secret sharing key with integrity verification and for identification of any tampering at any of the participating machine. So security of keeping the sensitive data on the cloud is higher than the previous algorithms. AEs default scheme is further using the POLY 1305 library functions to make the cipher test stronger and we also achieved in results that entropy of the proposed algorithm is higher than the default schemes. We also applied the testing of our implemented simulation on real test set and results are validated according to NSIT suites of cryptographic algorithms.

11 Future Scope Existing work can be extended by considering the addition with an unmasked constant, the addition with a masked variable, the multiplication by a scalar and the multiplication between two shared variables. But it will add on the complexity in the process and computation overhead will increase accordingly.

12 Research Significance The National Inpatient Sample (NIS) is the main willingly accessible inpatient care catalog in the United States. It covers data from about 8 million hospitals. Our setup is to create evidence to make health care safer, greater quality, more accessible, and

A Variant of Secret Sharing Protected with Poly-1305

119

to work in secure manner inside the constituent part of Healthiness, Human Services with added partners.

References 1. Amazon: Amazon elastic computes cloud. http://aws.amazon.com/ec2 (2013) 2. Salma, T.J.: A flexible distributed storage integrity auditing mechanism in cloud computing. In: Information Communication and Embedded System, pp. 283–287 Feb 2013 3. Wang, Q., Wang, C., Ren, K., Lou, W., Li, J.: Enabling Public Audibility and Data Dynamics for Storage Security in Cloud Computing, pp. 847–859, May 2011 4. Sravan Kumar, R., Saxena, A.: Data integrity proofs in cloud storage. In: 3rd International Conference on Communication Systems and Network, 4–8 Jan 2011 5. Hao, Z., Zhong, S., Yu, N.: A Privacy-Preserving Remote Data Integrity Checking Protocol with Data Dynamics and Public Verifiability, Sept 2011 6. Venkatesh, M., Sumalatha, M.R., Selvakumar, C.: Improving public audit ability. In: Data Possession in Data Storage for Cloud Computing, ICITISF, pp. 463–467 (2012) 7. Goubin, L., Martenelli, A.: Protecting AES with secret sharing scheme. In: International Association for Cryptographic Research. LNCS 6917, pp. 79–94 (2011) 8. Zhang, Z., Chee, Y.M., Ling, S., et al.: Threshold changeable secret sharing schemes revisited. In: Theoretical Computer Science, vol. 418, pp. 106–115 (2012) 9. Abbott, T., Lai, K., Lieberman, M., Price, E.: Browser-based attacks on Tor. In: Proceedings of the 7th International Conference on Privacy Enhancing Technologies (PET), pp. 184–199 (2007) 10. Aberer, K., Despotovic, Z.: Managing trust in a peer-2-peer information system. In: Proceedings of the 10th ACM International Conference on Information and Knowledge Management (CIKM), pp. 310–317 (2001) 11. Kumar, R., Verma, H.K., Dhir, R.: Analysis and design of protocol for enhanced threshold proxy signature scheme based on RSA for known signers. Wirel. Personal Commun.—Int. J. 80(3), 1281–1345 (2015). ISSN 0929-6212 (Print) 1572-834X (Online) 12. Kumar, R.: Cryptanalysis of protocol for enhanced threshold proxy signature scheme based on elliptic curve cryptography for known signers. Knowl. Comput. Appl. (2018). https://doi.org/ 10.1007/978. ISBN 978-981-10-6679-5. Springer Nature Singapore Pte Ltd. 13. Armbrust, M., Fox, A., Griffith, R., Joseph, A.D., Katz, R.H., Konwinski, A., Lee, G., Patterson, D.A., Rabkin, A., Stoica, I., Zaharia, M.: Above the clouds: a Berkeley view of cloud computing. Technical Report UCB/EECS-2009–28, U.C. Berkeley (2009) 14. Acharya, A., Raje, M.: MAPbox: using parameterized behavior classes to confine untrusted applications. In: Proceedings of the 9th USENIX Security Symposium (2000) 15. Aktug, I., Dam, M., Gurov, D.: Provably correct runtime monitoring. In: Proceedings of the 15th International Symposium on Formal Methods (FM), pp. 262–277 (2008) 16. Jianhang, Z., Hua, C.: Security Storage in the Cloud Computing: A RSA-Based Assumption Data Integrity Check without Original Data, pp. 143–147 (2010) 17. Patel, A.A., Jaya Nirmala, S., Mary Sarina, S.: Security and availability of data in the cloud. In: Advances in Computing and Inform Technology, AISC 176, pp. 255–261 (2012) 18. Abawajy, J.: Determining service trustworthiness in intercloud computing, environments. In: Proceedings of the International Symposium on Pervasive Systems, Algorithms, and Networks (I-SPAN), pp. 784–788 (2009) 19. Anderson, D.P., Cobb, J., Korpela, E., Lebofsky, M., Werthimer, D.: SETI@home: an experiment in public-resource computing. Commun. ACM (CACM) 45(11), 56–61 (2002)

Securing Bioinformatics Cloud for Big Data: Budding Buzzword or a Glance of the Future Bijeta Seth, Surjeet Dalal and Raman Kumar

Abstract Insight to utilize the Big data of Bioinformatics information generated by a paradigm; Cloud Computing is coming up as a guarantee to deal with big information storage and scrutiny challenges in the Bioinformatics field. Cloud computing is viewed to be a cost effectual technique to process and accumulate this immense quantity of data with parallel processing tools and carried as “Services” through the internet. Due to its fast and efficient performance for data processing on cloud clusters and easy to use environments, The Hadoop parallel programming framework is dominantly used. This document will be bearing in the direction of the productive course for economical Bioinformatics clouds for the Big data and also the challenges that would obstruct Bioinformatics Big data to take a stride towards the cloud. In this document, we state an outline of the applications of Bioinformatics clouds, merits, and limitations of the current research activity methods used for storing Big Data in Bioinformatics. The paper mentions how the existing dilemma can be addressed from the perspective of Cloud computing services in addition to Bioinformatics tools. For ensuring trust, a simulation comparing the trust values for different Cloud providers is being illustrated in Fog server. For Future enhancements, efforts are being made to build up an efficient cloud data storage system employing different Bioinformatics tools ensuring security so that various Healthcare organizations are benefited by this approach. Keywords Big data · Bioinformatics · Cloud computing · Secure cloud · Bioinformatics cloud tools · MapReduce · Fog B. Seth (B) · S. Dalal SRM University, Sonepat, Haryana, India e-mail: [email protected] S. Dalal e-mail: [email protected] R. Kumar Department of Computer Science and Engineering, I K Gujral Punjab Technical University, Kapurthala, Punjab, India e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_8

121

122

B. Seth et al.

1 Introduction Big data refer to voluminous magnitude of data—both structured and unstructured data that are so huge or intricate in nature to be solved by the traditional data processing application software. The existing outburst of information that is being generated is due to three chief reasons namely, Continuous information is being collected by various applications such as social media, mobile sensors, wireless sensors and other related devices, Storage space for collecting data is easily available nowadays at much cheaper rates than ever and Knowledge acquisition strategies have improved in a significant way with the help of more efficient information retrieval procedures. The major issues when addressing Big Data is allied with two chief characteristics are Huge volumes of information to be stored and organization and The procedure for delivering out the intent of these huge volumes of data, to determine valuable data and facts for prospect activities. With the exponential growth [1] in increasing data in Bioinformatics, demand for more storage space to be handled in further lithe and expenditure effectual mode is increasing day by day. Bioinformatics study in the post-genome period is extremely data-driven and supplementary integrative as the novel paradigms are functional such as NGS, mass geometry and imaging to recognize new organic insights. Nowadays, Big data imply scalability in addition to huge volumes of records. Cloud Computing presents a new approach to provide services to the user on pay per use basis and is highly cost-efficient and flexible. Cloud computing provides web services to the users through the Internet exploiting the concept of distributed computing. The customers pay for the services and resources they actually use. The manuscript is planned as follows. Section 2 gives a background of computational techniques used for analyzing Big data in Bioinformatics. Section 3 mentions the relationship between the three paradigms. Section 4 describes Bioinformatics clouds and various tool used for storage of Big data. Sections 5 and 6 mentions the pros and cons of the ways in which cloud provides solutions to the technique Related work constitute Sect. 5. Section 7 mentions the different ways how Bioinformatics and Life Sciences are using Big data and the cloud. Real life examples are discussed in Sect. 8. Section 9 describes the techniques for Bioinformatics applications. Section 10 discusses the challenges in Bioinformatics process and application of cloud computing. Section 11 describes the Results and Analysis being made. In the end, in Sect. 12, the conclusion showing the importance of the suggested solutions and their extensions is provided.

Securing Bioinformatics Cloud for Big Data …

123

2 Cloud Computing as Computational Amenities for Bioinformatics Applications The conventional platforms for Big data Analysis such as High-Performance Compute clusters had inadequate latent to adapt the computational environment. Bioinformatics applications or projects could be managed by Cluster Computing, GPU Computing, Xeon Phi., and cloud Computing [2]. We will discuss Cloud Computing here in detail. As Bioinformatics [3] are interdisciplinary in nature, it is promoting a novel development where Cloud computing effectively addresses Big data storage and Analysis issues. Virtualization is the key technology in Cloud computing to achieve dynamic, distributed and scalable systems to exchange great quantities of data and offer services on request and “pay as you go” technique over the internet. • Cloud Technologies They refer to the different runtimes [4] such as Hadoop [5], Dyrad, and other MapReduce frameworks, and also the different storage and communication frameworks such as Hadoop Distributed File System, Amazon S3. Figure 1 depicts Azure as the classical programming model, as a task processing pipeline approach where cloud instances (EC2/Azure compute) perform data processing and a queue of tasks for task scheduling. Windows Azure, provided by Microsoft, divides application running in clouds or data centers into roles. A role (web role, worker role, virtual machine role contains a specific set of code, such as a .NET assembly, and the environment in which that code runs.

Fig. 1 Azure cloud programming model

124

B. Seth et al.

Fig. 2 Computation of MapReduce

MapReduce [6] is a popular programming model for processing and performing data-intensive tasks on large data sets and performs transformations of data sets using two key functions: Map and Reduce. The MapReduce [7] actually is carried out in four phases: • Map function takes a key/value pair and computes a collection of this type of pairs: map :: (key1, value1) → arraylist(key2, value2) • An array is produced by combining sort phase and merge phase. • A key/value list is mapped to a list of values by a Reduce function. • r educe(key2, value2) → list(value3) The overall computation is demonstrated in Fig. 2. Hadoop is a very popular MapReduce runtime model developed by Apache. Dyrad is a functioning of MapReduce by Microsoft developed as data flow graphs to solve computational responsibilities.

3 Correlation Between Bioinformatics, Cloud Computing, and Big Data Cloud computing is promising its abundant applications ranging commencing business notion to one of the greatest emergent segments of the IT trade.

3.1 Big Data Manipulation in Clouds Cloud computing and big data [8, 9] are interconnected. Big data utilizes scattered storage tools based on cloud computing and processes the disseminated queries across diverse datasets delivering an ensuing set in an efficient manner. Cloud Computing

Securing Bioinformatics Cloud for Big Data …

125

Fig. 3 Cloud computing for big data

uses distributed data processing technique termed Hadoop using the virtualized technique for big data as depicted in Fig. 3. It permits distributed dispensation of storage by fault-tolerant parallelized MapReduce approach. Databases store voluminous data sourced from cloud and web in a fault tolerant manner in a cluster. Data visualization aids in decision making by analyzing the results. Examples of companies using Cloud platforms for Big Data are Google, Amazon, Microsoft, Cloudera etc.

3.2 Cloud Computing in Bioinformatics Convergence of cloud computing and Bioinformatics has created a spectacular product [10]. Pharma companies are gradually adopting cloud computing into their research practice. David Powers of Eli Lily mentioned emerging cloud computing and biotechnology saves a huge amount of time from weeks to three minutes by deploying an on-premise environment and earned huge amount of profits. Traditional methods of storing, processing are found to be ineffective as compared to the resource and cost saving methods of cc and big data [11] processing. They are showing benefits by curing diseases [12] faster and improving R&D in drug and treatments. Cloud computing has proved to save a lot of time and money of companies and benefited their research in dramatic ways by sharing research information by instantiating cloud tools on cloud computing infrastructures. Example: Galaxy instantiated on the Amazon Elastic Compute Cloud (EC2). Again, Hadoop meets the Bioinformatics [2] requirements for distributed data handing out of huge data sets, allows data scaling (HDFS, HBase) and MapReduce enables parallelized examination in a fault tolerant manner.

126

B. Seth et al.

For the Bioinformatics domain, Big Data relies on the four paradigms. Initially, the volume of data requires distributed techniques for information query and retrieval. For instance, the ProteomicsDB includes 92% (18,097 out of 19,629) of human genes recorded in the Swiss-Prot list having a gigantic records quantity of 5.17 TB. Electronic Health Records (EHRs) have tripled in healthcare organizations to 44% from 2009 to 2012 according to HITECH Act [13] to enhance the healthcare services and increase the research opportunities. The second ‘V’; the variety of data sources, i.e. data structures and data types, causes prevention of next relational structures and utilizes unstructured data(OHR, Medical images) providing many opportunities to formulate new investigations [13]. Example, Genomics, Proteomics, and Metabolism generate “OMICS” at all levels of Cellular components. Thirdly, the third ‘V’, the velocity signifies the pace at which the information is generated i.e. producing and processing the data. Example: Billions of DNA sequence data are being generated by sequencing technologies every day at a comparatively small price. This “need for speed” is required to provide Bioinformatics [14] researchers with timesaving tools for discovering new patterns of medical data. This implies, scientists and researchers have to store, analyze, compare and share large data sets which are not a trivial task. Finally, the fourth ‘V’, the Veracity of data determines the uncertainty of data. The customer’s data must remain consolidated, cleansed, consistent and current to make the right decisions. Because of the 4 V’s of Big data, it is essential to building up pioneering solutions that incorporate dissimilar approaches in a direction to generate lithe and pliable systems.

4 Storage of Big Data in Bioinformatics Clouds The traditional method for Bioinformatics analysis involved three major steps: 1. Download data from public sites(e.g.: NCBI, Ensemble) 2. Install software kit locally 3. Running investigation on in-house workstation resources. Albeit, cloud computing holds a vast guarantee in successfully addressing big data storage and Bioinformatics trouble, Public access to biological data and Bioinformatics tools must be made available for efficient and low-cost storage and analysis and carried as “Services” through the internet viz DaaS, SaaS, IaaS, and PaaS. Bioinformatics cloud computing has changed the momentum of research projects by lowering the costs and viability of cloud computing in pharmaceutical R&D. Cloud-based Bioinformatics [15] platforms address the fundamental demands for creating a flexible, scientific environment, facilitating data processing and a multitude of advantages as demonstrated by Omics technologies; helping to support both government and scientific mandates. • Cloud-based SaaS solutions accomplish diverse Bioinformatics [16] responsibilities, e.g.: mapping applications, sequence alignment, and gene expression analysis

Securing Bioinformatics Cloud for Big Data …

127

• Cloud scale rsources(sequence mapping, alignment, assembly, sequence analysis) • Cloud based applications for NGS,data analysis • Pipeline architecture • Analysis platform • Programming environment • Web servers and databases

• Virtualized machines • Virtualized resources

• Public datasets like AWS • Biological databases

Fig. 4 Illustration of bioinformatics clouds [18]

using cloud-based tools. The different genome sequences methods use Hadoop implementation of MapReduce (like eCEO, StormSeq, Crossbow), short-read aligner (CloudAligner, CloudBurst), variant annotation (VAT) and RNA-sequence (i.e. FX, Myrna) to parallelize completion using numerous workstations and maintain access by a comprehensible web interface. • Cloud-based PaaS solutions allow customers to adapt the operation of Bioinformatics applications and to preserve absolute control over their instances along with allied data. CloudMan, GalaxyCloud, and Eoulsan are the main examples of PaaS solutions for Bioinformatics applications. Azure [17], GoogleApp Engine is referred as PaaS (Fig. 4). • Cloud-based IaaS solutions: This service model includes virtualized servers with the precise computational potential and storage. Various deployed resources viz storage, OS, and Bioinformatics applications are controlled by the user. IaaS solutions used to process genomics [19] and phenotypic data, include Bionimbus, CloVR [20], and CloudBioLinux. Amazon EC2 is seen as IaaS in Bioinformatics. • Cloud-based DaaS service allows encompassing efficient data that are available from an ample variety of allied devices on the network by providing information in a virtual environment hosted by a cloud (Table 1). Because of the above advantages, cloud computing services are providing their services from medium to large scale solutions by a powerful infrastructure at an affordable cost. Below are the various Bioinformatics cloud tools available for various services in a tabular form.

128

B. Seth et al.

Table 1 Bioinformatics cloud tools [21, 19] Resource

Year

Description

Application tools

Software as a service CloudBlast

2008

The software implementation of NCBI Blast using Apache Hadoop integrates virtualization, virtual Network technologies using ViNe used for Grid computing and parallel techniques for bioinformatics applications

Hadoop, Blast, ViNe

CloudAligner

2011

Quick and complete feature MapReduce based device for sequence analysis

CloudBurst, MapReduce, Amazon EMR

CloudBurst

2009

Read mapping Algorithm using MapReduce for mapping single and NGS data and are faster than short-read mapping software RMAR Parallel processing in MapReduce done by Seed and Extent algorithm

MapReduce, Amazon EC2

CrossBow

2009

Programs used for Human Genome alignment and Single Nucleotide Polymorphism (SNP) detection Long run sequential data processing using CC resources (Amazon EC2) and MapReduce easily completed in lesser time

Hadoop, Bowtie, SOAPsnp, Amazon EC2

FX

2012

An RNA Sequential investigation device intended for the cloud

Hadoop, Amazon EC2

Myrna

2010

Cloud-scale RNA sequencing differential idiom investigation means

Hadoop, AmazonEMR, HapMap

eCEO

2011

Efficient and flexible Epistatic computing model for identification of interactions in genome-wide study used for load balancing across processing nodes

Amazon EC2

PeakRanger

2011

Chip-sequential data offered improved Resolution, excellent sensitivity, and specificity and above average spatial accuracy in terms of identifying the precise location by peak enabled caller

Amazon EC2

RSD

2012

Amazon EC2 uses Reciprocal Smallest Distance algorithm for ortholog detection

DIOPT tools, Python

VAT

2011

Variant Annotation Tool annotates variants from numerous personal genomes at the transcript level Two external libraries required before installing VAT: libBIOS and GD library

VCFtools

YunBe

2011

Open source and freely accessible pathway used for gene set analysis

Amazon Elastic cloud, MapReduce (continued)

Securing Bioinformatics Cloud for Big Data …

129

Table 1 (continued) Resource

Year

Description

Application tools

Rainbow

2011

Enhancement of crossbow to detect SNP in whole genome sequencing (WGS) on an Amazon Hadoop cluster Better than crossbow in four ways (1) Support for BAM input files (2) Data pre-processor to deal with large FASTQ files quickly by splitting them into smaller pieces (3) Check cluster nodes (4) SOAP SNP aggregator to support. Genomic-wide association studies (GWAS)

Crossbow, Bowtie, SOAPsnp, Picard, Perl, MapReduce

AzureBlast

2010

Parallel BLAST (Basic Load Alignment Search Tool) for Windows Azure used for studying science applications. Finds local similarity between sequences

Azure, BLAST

Platform as a service EoulSan

2012

Hadoop execution of the MapReduce algorithm devoted to outstanding throughput sequencing information analysis

Hadoop, MapReduce

GalaxyCloud

2010

Cloudscale for large-scale data analysis Reproducible Research System (RSR) is applied to support reproducible computational experiments for reading and repeating user workflows It works with Sun Grid Engine (SGE) or portable batch systems(PBS) to support distributed workloads in the cluster

Python, Web server, SQL database

Galaxy CloudMan

2010

Galaxy CloudMan console manager computes cluster on the cloud to increase/decrease the cluster size Cluster jobs are controlled by SunGrid Engine (SGE) Amazon Elastic Block Storage (EBS) keeps data after shutting down cloud instances

Amazon EC2, BioLinux, Galaxy

CloudBioLinux 2012

Community project VM designed for biologists and programmers for the genomics community

VM, Amazon EC2, Eucalyptus, VirtualBox

CloVR

An application for pushbutton automated sequence analysis utilizing cc resources

VM, VirtualBox, VMware

Infrastructure as a service

2011

Data as a service AWS public datasets

Cloud-based records of Genbank, Ensemble, Influenza virus

130

B. Seth et al.

5 Advantages of Bioinformatics Clouds Cloud computing in Bioinformatics provides the organizations with the power and flexibility required for research and business world [11]. Other significant advantages as mentioned in Fig. 5 include: • Compliance and access control: Cloud Computing allows biotech companies to store, access, and share data across multiple companies with proper procedures for access and authentication control like electronic signatures. Control panel GRC in SAP HANA reduces administrative overhead and automates 60% of governance, risk and compliance control. • Big storage space: Cloud Infrastructure as a Service is very affordable because of “pay as you go” feature, elastic and easily deliverable nature. • Data transfer: Data can be stored, accessed and transferred easily to the cloud and thereby reduces the chance of unavailability because of hardware failures. • Real-time reporting: SAP HANA LIVE and SAP S/YHANA and other companies easily face internal and external challenges by using granular control, decision making and quality control feature provided by cloud computing. • Big data recovery: great computational power, performance enhancement, increased speed is provided by combining cloud computing with biotech allowing big data analysis. Example: speed of SAP HANA is increasing manifolds, Human Genome project could be completed in mere 26 h instead of decades. Without a shadow of a doubt, the ongoing use of big data and cloud computing will no doubt accelerate advances in Bioinformatics.

Compliance and access control

• Cloud Computing allows biotech companies to store, access, and share data across multiple companies with proper procedures for access and authentication control like electronic signatures.

Big storage space

• Cloud Infrastructure as a Service is very affordable because of “pay as you go” feature, elastic and easily deliverable nature

Data transfer

• Data can be stored, accessed and transferred easily to the cloud and thereby reduces the chance of unavailability because of hardware failures

Big data recovery

Real time reporting

• The mergeed technologies provides great computational power, performance enhancement, increased speed. Example: speed of SAP HANA is increasing manifolds, Human Genome project could be completed in mere 26 hours instead of decades. SAP HANA LIVE and SAP S/YHANA and other companies easily face internal and external challenges by using granular control, decision making and quality control feature provided by cloud computing

Fig. 5 Advantages of bioinformatics clouds

Securing Bioinformatics Cloud for Big Data …

131

6 Disadvantages of Bioinformatics Clouds Big data in Bioinformatics countenance different implementation challenges [18] which hinder clients to move their data to the cloud regardless of several benefits. Data center have restricted facility to accumulate and backup massive biological databases (Terabytes, Petabytes). Performance demanding examination of tremendously large Bioinformatics databases either is not available or is elongated to complete. Programming Hadoop demands a highly sound Java proficiency to develop parallelized programs. As a solution, Software libraries such as Hive are added as an “SQL” resembling interface that will produce parallelized Hadoop jobs in the background. Also, Java programming is being wrapped in Python, creating a more lightweight scripting language. The rate of data transfer, the technique of transmitting scale of genomic data over the Internet or any other form of communication media takes prolonged periods of time. Fundamental aspects of data security need to be addressed before widespread adoption of cloud-based applications to occur, including Encryption mechanisms, the vulnerabilities of Internet-based customer access interfaces, replicated in the instance of disaster recovery and inadvertent data access. Service interoperability of cloud service (the ease with which data can be transmitted) is difficult for customers to drift from individual source to a different or move back information and services. Data privacy legislation, as well as permissible possession pertaining to records stored among global zones, is difficult to maintain. Standardization of reporting and summarization of results is a trouble; there is a need to build up improved analytics and virtualization technologies for developing programmer friendly management interfaces. Example Hadoop with no font and visualization is difficult to set, maintain and use. Of course, it’s not all roses and daffodils, and above are some of the disadvantages to adopting Bioinformatics cloud. Cloud computing security issues will probably prolong to hamper its utility to pre-competitive or non-functional data.

7 Related Work Radical developments are brought largely by the advent of cloud computing and Bioinformatics. This section presents an outline of existing review articles allied to Bioinformatics cloud and Big data. Troup [6] proposed Metamorphism testing instead of Oracle testing. The author discussed how a cloud-based metamorphic testing can be applied to a genome sequencing data using cloud-based resources. Dowlin et al. [21] described challenges of hosting, sharing and computation in biological data science with large datasets. Preserving Privacy of sensitive information of personal health and genomic data is a significant concern. Homomorphic Encryption computing data without revealing the decryption keys were discussed. The paper introduced homomorphic Simple Encrypted Arithmetic Library (SEAL) publically for Bioinformatics and genomics. Misirli [22] presented the Protocol for Linking External

132

B. Seth et al.

nodes (POLEN) as a cloud-based scheme facilitating synthetic biology design workflows operating asynchronously. POLEN can be used for designing and implementing biological systems to guarantee uniformity of information crossways dispersed repositories. Moghaddasi [23] intended to study the handling of cloud computing services and surveyed abundant documents and concluded that the applications of cloud computing in health systems, counting telemedicine, medicinal imaging, community and private health, remedial judgment support system serve as a means to analyze gene sequences and files huge biological information. The cloud services in the health system are basically available as E-health services and Bioinformatics. Kumar [24] mentioned the requirement to set up and test cloud programmable environment where resources can be automatically and dynamically matched with the requirements with minimal concern to customers of resource allocation. The paper employed MapReduce scheme with indexing to fulfill user’s requirements in the bio-analytical field. Calabrese [25] discussed major issues and troubles for storage and analysis of patient’s data allied to use of cloud-based Bioinformatics solutions were mentioned. Analysis of cloud-based Bioinformatics tools and platforms including SaaS, PaaS, DaaS, and IaaS were mentioned. Prachi [18] mentioned that cloud computing has a potential in Biotech industry. Parallel DNA Sequencing generates an immense quantity of data. Genomics revolution resulted in Metagenomics [10] and lead to the sequence-based investigation of the Microbiome (microbial genome). Lee introduced the parallel application of Bioinformatics and studied diverse projects concerning to reduction of data analysis time and improving usability with their approaches. Hadoop implementation and workflow toolkits were suggested to address parallel data processing tools HDFS and MapReduce for an easy to work virtual environment for Terabyte-scale data analysis in Bioinformatics parallel problems. Celesti [26] mentioned that Next Generation Sequencing (NGS) allowed sequencing DNA at a relatively low cost and higher speed and had given a new dimension to the learning of Genomics and Molecular Biology. The paper highlighted the chief NGS Cloud-based solutions along with the challenges and advantages in scientific laboratories. DNA data generated by NGS [26] is an exemplar of Big data which is difficult to be handled by traditional systems and thus cloud computing solutions are required to handle it. Guan [27] developed a cloud-based framework able to work on public and private clouds to deploy the bioinformatics tools on cloud virtualized platform based on Hadoop. The deployed tools were tested on Providence University cloud platform. The proposed framework is said to be efficiently deployed on well-known bioinformatics tools as web services. Afgan and Krampis [28] suggested that complex bioinformatics applications could be encapsulated within virtual machines along with the software. The scalability of cloud computing can be fully utilized through virtualized compute clusters and bioinformatics pipelines. The author mentioned multi-clouds for presenting automated entire system using web applications on any cloud. Lukas [29] elaborated MapReduce parallel programming and Apache framework as the novel solution for efficient data processing and storage in the field of Bioinformatics. Albeit, the potential of MapReduce is evident, still, there are two prerequisites for their adaption and utilization firstly, intricate workflows derived from several Bioin-

Securing Bioinformatics Cloud for Big Data …

133

formatics MapReduce tools hiding the technical particulars of execution and secondly the computing infrastructure for data processing. The paper presented integrated system allowing liberation of Bioinformatics MapReduce applications in the cloud by combining Cloudgene and CloudMan for delivering application execution environments. Shanahan [17] discussed the usability of resources on PaaS based Microsoft Azure for Bioinformatics for analyzing Microarray expression data deposited in the public repository, Array Express. The results illustrating a comparison between load time and run time calculations on Azure with the locally run machine were made in C# language. The author mentioned that a considerable set of steady Bioinformatics Software such as EMBOSS or BioLinux might subsist using Azure and its PaaS. Coutinho [30] merged GraspCC, proving next to optimal assessment of a number of Virtual machines for applications and SciCumulus, a cloud-based equivalent appliance for scientific workflows together to assess the indispensable quantity of virtual machines for Bioinformatics cloud workflows. Marx [4] introduced the parallel application of Bioinformatics and studied different projects. Hadoop implementation and workflow toolkits were recommended to tackle data processing tools HDFS and MapReduce. Hsu [19] aimed to foster the dissemination of high-quality research and practice allied to cloud and Bioinformatics considering genomics and drug design on the cloud. Lin [31] mentioned that biomedical data collected from various sources has increased manifolds and Cloud computing is an alternative to storage and computing facilities. Cloud-based resources and applications in Bioinformatics were summarized along with the variant cloud technologies which could be beneficial to researchers. Driscoll [5] mentioned an outline of Big data and cloud computing technologies and their challenges along with their association with Biology’s gigantic information sets. Present usage of Hadoop within Bioinformatics community was provided. Nemede (2013) mentioned a security model for flexible use by consumers and explained various cloud-based services and techniques for implementation of Big Data of bioinformatics clouds. Several issues and solutions of Bioinformatics big data were mentioned. Gabriel [32] focussed on job development intricacy in a federated cloud infrastructure named BioNimbuz for executing Bioinformatics applications. The author presented a scheduling algorithm based on Load Balancing Art Colony called ACOsched to find the best cloud in the federation for task execution and showed momentous enhancement in the make-span instance of Bowtie mapping tool as compared to round-robin algorithm called DynamicAHP.

8 Different Ways How Bioinformatics and Life Sciences Are Using Big Data and the Cloud This exertion reviews chief intellectual and manufacturing cloud based Bioinformatics solutions created in latest years along with their underlying issues and challenges related to the storage and investigation of patient data [13]. Genomics by using big data and the cloud for the Bioinformatics companies have enabled the researchers

134

B. Seth et al.

to analyze and match gene variants with specific diseases to find the treatment of cures. The Human Genome [33] Project is the best example of decisive the series of chemical base pairs which constitute the human DNA and recognizes and maps the genes of the human genome. Smart treatment through smart devices involving health monitors, the widely used connected smart device stores and processes, human health records in the clouds and helps in the better individual treatment and preventive care. Early diagnosis of problems prevents worsening of symptoms and conditions. Drug discovery is the merge of big data and cloud computing has proven to drastically transform the discovery of new drugs and medicines to cure diseases by utilizing the resource sharing characteristics of clouds. Optimal patient treatment and management through data sharing feature of the cloud have enabled the researchers to collect, share, analyze and correlate the critical trends and patterns of a specified group of people for medicine from varied geographical areas. The human microbiome is a collective micro-organism can be quickly and in detail them to by researchers allowing for curing a myriad of human gastrointestinal diseases. Proteomics [34] refers to the quick identification, prediction, and modeling of an array of proteins has made feasible the development of cure and drugs by different sharing of information brought on by cloud and Big Data. Crowd-sourcing involving cloud-based innovation allows funding of projects using the social web to answer Healthcare questions easily by many biotechnology companies. Clinical Informatics is used for making intelligent decisions using collected patients data. Big data tools/technologies for Clinical Informatics used are (a) Data storage and retrieval of EHR (EEG collection) is done using Hadoop based Cloudware to accumulate medical information and enables to develop a network-based interface for authentic retrieval. (b) Interactive information retrieval for enhancing data sharing between researchers can be done by using an application architecture based on cloud approach and Hadoop/HBase infrastructure. (c) Data Security: can be provided by using MedCloud which uses Hadoop ecosystem when accessing patient’s data and Home diagnosis to address privacy challenges and secure record retrieval for observance issues with HIPAA. (d) Data Analysis: PARAMO may be used to support invention and reclaim of clinical data analysis for the dissimilar modeling purpose. Imaging Informatics: Medical image data and workflows are efficiently produced after the integration of Big data with the cloud platform. Image Technologies/tools are used for Data storage and retrieval involving DICOM protocols (digital image communication in medicine) are used with PACS to deliver images, Data and workflow sharing through SuperPACS which enables a radiology group to serve multiple sites and Data Analysis through Hadoop-GIS is used as an efficient and cost-effective parallel system. Public Health Informatics: Infectious disease outbreaks can be predicted and monitored using Big data techniques. There are three core functionalities: Assessment: aims at gathering and analyzing information to the trail and scrutinize public health status, Policy development: depends on assessment phase, and Assurance: validates services offered through policy development have achieved target goals for increasing public health outcomes.

Securing Bioinformatics Cloud for Big Data …

135

The above discussion illustrates how Bioinformatics Clouds has become a significant paradigm for bringing the benefits of cloud computing to big data and Bioinformatics.

9 Real Life Examples Several markets are clawing for a share of a market analysis, emerging Cloud computing and Bioinformatics to take advantage of their operations.IBM Watson Health cloud service along with Teva Pharma in 2005 merged together aiming to help actors in healthcare to surface new insights from PHR and select small molecules for drug development. Sage biometric used Watson Health in 2015 to advise and guide clinical trial planning. An alphabet created by Google includes Healthcare subsidiaries of Calico, life sciences, Google venture and Google X and maintains information ranging from lenses to genes. AWS works in collaboration with Pfizer, Siemens, and Novartis to scale up experiments and improve productivity levels. Pfizer and their Biotherapeutics and bio-innovation center used AWS to model antibody docking, which shortened the process from 2/3 days to just 2/3 h. Eli Lily using Cloud computing launched a 64-bit machine cluster functioning on Bioinformatics sequence information, finished the entire work and shut it down in 20 min from a 12-week course. Several cancer organizations are working on sharing data with certain genomes enabling doctors to effectively cure treatments and finding how cancer patients react to their drug treatments. Eric Dishman, Director of Proactive Health Research at Intel shared his personal information regarding his treatment of Kidney cancer. Asclay Christenser in his book on healthcare [13] disruption mentioned the power of cloud computing to transform the intuitive medicinal system to precision medicine. Nancy Brown, CEO of American Heart Association mentioned cloud to be a game changer in the curing of diseases. Above scenarios depict the growing usage of Cloud computing in Bioinformatics market.

10 Basic Concepts in Bioinformatics Process and Application of Cloud Computing The broad classification of the various problems and the associated Bioinformatics tasks, their requirements, and the ways in which cloud computing techniques are applied to solving them can be categorized into 4 types [31]: • Data storage and retrieval: It includes the following: – The Genome sequence mapping needs to be mapped to specific reference genomes i.e. determine the relative positions of genes on a reference chromosome and mea-

136

–

–

–

–

B. Seth et al.

sure the distance between them with the aim to find specific nucleotides called Single Nucleotide Polymorphisms (SNP). The Algorithms used can be CloudAligner [35], CloudBurst, BlastReduce, and SEAL. Nguyen [35] built a Hadoop MapReduce based framework, CloudAligner to deal with long sequences for reading mapping and accepts two input files: the Read file (partitioned into denied size chunks and allocated to mappers) and the Reference file. It doesn’t have the Reduce phase of the traditional MapReduce model. Schatz et al. [36] designed a highly sensitive modified RMAP, parallel seed-and-extend read mapping algorithm called CloudBurst to optimize mapping single end next-generation sequence data to reference genomes. It is a MapReduce based (map, sort and reduce phases) read mapping algorithm run on multiple machines in parallel with Hadoop. [37] introduced SEAL as a group of disseminated applications for aligning, manipulating and analyzing short DNA [38] sequence reads. BlastReduce was introduced as an analogous short DNA sequence read mapping algorithm optimised for aligning sequence information. Genomic sequence analysis for recognizing Single Nucleotide Polymorphism (SNPs) [39] from sequencing statistics. The Algorithms used can be CrossBow, Contrail, CloudBrush. The core of Genomics remains sequencing applications including sequencing new species, profiling high abundance of each taxonomic unit in an environmental sample, arranging DNA sequences, RNA or protein to recognize regions of similar functions and constructing the 2D/3D structure of RNA or protein based on 1D sequence. Gunarathne [40] presented two adjustable classical assembly program Cap3 to align and merge sequence fragments to determine the whole genome sequencing by utilizing Apache Hadoop and Microsoft DyradLINQ. Langmead [39] introduced CrossBow, a Hadoop based software tool which performed alignment ad SNP detection for manifold whole human datasets per day. RNA sequence analysis used to find differential expressions from billions RNA sequence reads. A tool used can be Myrna, FX, and Eoulsan. Ben [41] offered a cloud tool for computing differential gene expression in large RNA-Seq datasets called Myrna. Hong [42] generated FX for RNA sequence analysis device for the inference of gene expression levels and genomic variant calling. Jourdren [43] proposed Eoulsan as an incorporated and bendable result for RNA sequence data examination of the differential term [5]. Search engine Implementation is used for dealing out huge peptide and spectra databases that support the scalable penetrating of enormous spectrometry records. Tools used can be Hydra, CloudBurst. Lewis [44] described Hydra as a scalable proteomic searching tool utilizing the distributed Hadoop structure. Matsunaga [45] proposed CloudBlast as a scalable BLAST in the cloud. Sequence file management helps Apache HBase databases to assist Bioinformatics researcher’s access great scale—whole genome datasets. The tools used are Seqware, GATK, Hadoop BAM. Niemenmaa [46] invented the Hadoop-BAM as a novel library for scalable manipulation of aligned next-generation sequencing data. Merriman [47] founded a query engine SeqWare as a tool-set used for nextgeneration genome sequencing technologies which include a LIMS, Pipeline and

Securing Bioinformatics Cloud for Big Data …

137

Query Engine. McKenna [48] discovered the Genome Analysis Toolkit, GATK, A gene analysis tool-kit for next-generation resequencing data. • Error Identification: tools developed to identify errors in sequencing data can be SAMQA, ART, CloudRS. • Data analysis: tools include Genome Analysis Toolkit (GATK) [48], BlueSNP, Myrna, Eoulsan. Huang [49] found an algorithm for large genotype-phenotype datasets for performing computational intensive analysis, BlueSNP. • Platform Integration deployment: New techniques are required to combine existing big data technologies with easy to use operations. Tools like SeqPig, CloVR, CloudBiolinux are used for the same. Angiuoli et al. [50] implemented CloVR as a single portable VM runs on local computer resources and requires minimum installation. Krampis [51] configured CloudBioLinux as a virtual machine with over 135 bioinformatics packages.

11 Challenges in Adoption of Bioinformatics Clouds Medicinal bioinformatics, in reality, is frequently fretful by perceptive and costly data such as projects causative to computer-aided prescription or in scenarios similar to hospitals. The sharing of data raises the intricacy and involves data transmit through numerous networks. There are several scientific obstacles to an acceptance, intensification, strategy, and commercial obstacles for shifting to cloud computing. Some of the significant barriers to cloud computing responsible for its slow adoption in healthcare are explained with security in greater detail [52, 53]: • Privacy and Security challenges: data stored in a cloud may contain personal, private and confidential information which requires being protected from disclosure or misuse. A service level agreement among cloud service provider and covered entities including institutions, organizations or persons transmitting the health information electronically need to be established under HIPAA regulations. Whenever PHI data flows over open networks, it must be encrypted. The HIPAA compliance requires recognition and authentication of users along with a complete authorized privilege and role-based admittance control like passwords to validate the uniqueness of users seeking permissions to access the PHI information. • Workflow challenges: From paper work and data entry system to cloud model, new training and new skill sets need to be planned. Electronic medical record (EMR) should be emphasized. • Reliability and performance: Disaster recovery is a component of service reliability is difficult to develop as most SLA doesn’t provide satisfactory guarantees in case of a service outage due to a calamity. Another factor is a performance which is affected by a mismatch in types of internet connections which vary at different speeds.

138

B. Seth et al.

• Integration and interoperability: The Standard Development Organization (SDO) develops specification and principles to sustain healthcare information, exchange of information and systems integration which is difficult to maintain. • Data portability and mobility of records: The movement of data between healthcare organization and cloud vendors requires no disruption to data. The migration to a new cloud service provider (CSP) in case if any CSP goes out of service becomes a serious issue. A sensible substitute is to actually transport hard drives to the cloud vendors. • Speed: Depending on information extent and net hustle, transfers towards and from the cloud can cause a considerable hurdle. By using cloud computing, faster and accurate access to all information for healthcare service can be made by using high-speed backbones like Internet2, Janet etc. High-speed Internet backbones are required to connect for transferring large data sets from and to the clouds. • Computational Biology Problems: A foremost dilemma in this pasture is epistemological [54] which have restricted its utilization in research exertion. • Redesign of Parallel frameworks: Cloud computing frequently involves redesigning applications for parallel frameworks like Hadoop which requires proficiency and is prolonged. Latest parallel programming systems through their encoding abstractions, such as DyradLINQ, facilitate various cases [38].

11.1 Security and Privacy The notion of handling important data to a third party is worrisome and the customers need to be cautious in understanding the risks of data breaches. Security and privacy need to be addressed with respect to Bioinformatics data also. This section briefly explains the related work to security. Heitor et al. [55] described the security problem occurring in the federated clouds which prevent users from adopting federated clouds. The author focused on the problem of access control in federated clouds by using the attribute-based model to decide whether the user is authorized to access a given resource. The performance evaluation was performed using a federated cloud specifically designed for bioinformatics applications called BioNimbuz and effectively restricted access to resources and other applications. Yamamota and Oguchi [56] proposed a client-server system using Fully Homomorphic encryption for string search of genome sequences in Bioinformatics research. The author aimed to improve the query and calculation algorithm using a decentralized system on the server side of a string match system for the genome operating in a cloud environment. Namasudra et al. [57] proposed efficient data accessing technique using 512-bit Deoxyribonucleic Acid (DNA) based key series for recuperating the data scrutiny against the collusion attack, MITM attack etc. The author mentioned data security as biggest concern in big data other than data storage, high data access time, elevated information search time etc. The proposed scheme appeared to be proficient theoretically and experimentally than other schemes. Yang et al. [6] examined software scalability and validity as two important aspects of big data bioinformatics analysis.

Securing Bioinformatics Cloud for Big Data …

139

The author explained the use of big data programming frameworks such as MapReduce and Spark and other software testing techniques based on multiple executions such as metamorphic testing as an effective measure for Bioinformatics quality assurance. Shi et al. [52] prescribed the combination of Cloud computing and the Hadoop MapReduce parallel programming for analyzing Next Generation Sequencing (NGS) data in a cost-effective and timely fashion. However Hadoop tuning guiding principle are considered to be outdated or too common for Bioinformatics Applications. The author aimed to select Hadoop key parameters for NGS data analysis to avoid resource wastage. The paper resented k-mer counting and achieved 4* speedup as compared to original performance parameters. Silva et al. [58] developed a cloudbased system to analyze next-generation sequencing data with pros like on-demand scalability, ease of use, access control, and security for forensic genetic applications. Relational database stored the resulting data and multi-layered architecture applied to NGS system was used for data analysis, storage and multi-user distributed connectivity. Abdulunabi et al. [59] described the Nationwide Health Information Exchange (NHIE) as a means of providing global connectivity and interoperability for exchanging patient information among healthcare providers. However, security has always troubled the government from successfully deploying NHIE. The paper proposed a new approach by leveraging the notion of mobile individual health records, mPHR (PHRs installed on smartphones) as a reliable and indispensable source for information exchange between healthcare providers and mPHRs through special interfaces. The new framework with overall design specifications of components and architecture were provided along with the proof-of-concept prototype to validate the notion. Goyat and Jain [60] aimed to design a secure cloud storage and data distribution platform for public cloud and thus surveyed a number of traditional and modern cryptographic techniques. The author proposed the efficient and cost-effective DNA based cryptographic algorithm for security developed using the substitution and other operator’s implementation. The security enhancements are made by allowing only four symbols to be used for transforming the message to cipher-text. Java language was used for implementing the data exchange and sharing and OpenShift cloud was used as a public cloud for deployment. Space and time complexity was found to be effective and low resource consuming in terms of data security in the clouds. Nepal [61] mentioned that different workflow environments and tools like Taverna and Galaxy have evolved to be integrated with cloud tools like CloudMap, CloudMan etc. for service delivery and data integration. Albeit, Cloud security causes severe confront concerning the acceptance of cloud infrastructures like Health sector where genomics data and medicine sector need much attention. The paper described an architecture termed TruXy for concerted bioinformatics research in endocrine genomics and is experimented for dealing out of exome data sets on individuals with genetic disorders. The author compared the performance of TruXy with standalone workflow tool and mentioned its success for handling big data security challenges. MeteAkgun et al. [53] mentioned the reason for the huge amount of sensitive genomic data as advances in genome sequencing and privacy challenge occurring in them. The survey had categorized pre-existing problems and their corresponding solutions in a convenient and understandable manner. The author believed that their

140

B. Seth et al.

classification of genome associated privacy problems arising with genome data processing will pave the link between real-life problems and methods. Siddaramappa and Ramesh [62] discussed cryptography and bioinformatics algorithms for providing a path for secure channels. The paper presented a new method of Deoxyribonucleic acid and RNA as keys for encryption/decryption technique to hide data from intruders. The proposed system was considered to be providing high levels of security as compared to other algorithms by generating a random sequence of deoxyribonucleic acid and converting them into a binary representation. Liu et al. [63] proposed a method for deploying Galaxy on Amazon cloud-based Bioinformatics workflow for NGS analysis and integrated it with Globus Transfer for high performance and reliable data transfer. Auto-scaling and parallel computing were achieved by integrating Galaxy with HTCondor scheduler. Performance evaluation was presented by using two Bioinformatics workflow use cases to validate the feasibility of the proposed approach for efficient data sharing and retrieval. Razick [64] mentioned the importance of protected verification and security process to guarantee firm access policies and reliability of data in case of biomedical [31] applications in a distributed scenario. Rodrigues [65] mentioned the Role bases access, third party certificates, logging access to data and computer network security are some of the techniques used to address security concerns. Alex [66] discussed the concept of cloud computing and its role in healthcare with four main aspects to assess the opportunities and challenges of cloud computing. The deliberate forecast could benefit healthcare organization when migrating from traditional to cloud-based health care services. Ney and Koscher [67] mentioned that big data revolution in genomics sciences has led to the creation of Bioinformatics tools. The paper demonstrated the sequencing and processing of DNA could allow an attacker to allow remote code execution to introduce a vulnerability. Sequencing community phenomena caused information leakage and also showed poor security. The document developed an outline and strategy to preserve safety and confidentiality in DNA synthesis, sequencing, and processing.

11.2 State of the Art and Proposed Solution We have observed that the major issue to store and move Bioinformatics data in the cloud is its security. Security of data has always been a prominent concern in any organization. Hamid Abdulaziz et al. [68] aimed to provide a secure healthcare private data in the cloud using a Fog computing paradigm. Electronic Medical Records containing huge multimedia bug data including X-rays, ultrasounds, MRI reports etc. are required to be efficiently accessed by both healthcare professionals and patients. A tri-party one round authenticated key agreement protocol based on Bilinear pairing cryptography generating session key among the participants were generated for secure communication. Gonzalez et al. [69] mentioned the invention of Fog computing in 2012 and pointed cloudlet to be a part of edge computing useful to mobile networks and fog notion as a subset of edge computing applied to the Inter-

Securing Bioinformatics Cloud for Big Data …

141

net of Things. The status and potential aspects of edge computing were discussed. Stojmenovic and Wen [70] mentioned Fog computing [71] as an extension of Cloud Computing and services to the edge of the network. The motivation behind Fog computing and its advantages and applications in real life like Smart Grid, smart Traffic lights in vehicular networks etc. were discussed. Security and privacy issues were discussed elaborating Man in the middle attack by measuring its CPU and memory consumption on a Fog device. When considering the data outsourcing to a cloud, standard encryption techniques can be used for security of operations and storage of data. The data before sending to the cloud is encrypted which increases the security level. Every time an operation is to be performed on data, it needs to be decrypted first and the client is required to give private key each time to the cloud provider. Thus affecting the confidentiality of data as the server is not authorized to perform any operation. A new kind of encryption is proposed for users with different access structures. Data is stored in encrypted form and decryption will take place only in sections by the users based on their identity. Thereby, dependency on storage server for preventing unauthorized data access is reduced. Secondly, usage of digital certificates for encrypted data makes it difficult for malicious users to open it. We have used Fog computing which is an extremely virtualized policy providing compute, storage and networking services among end devices and is well thought-out as an extension of Cloud computing as an “Edge of the Network” [69].

12 Results and Discussion For security, ensuring trust, Fog can be used to provide an easy access and service compatible platform. Fog (https://gist.github.com/kordless/6922896) is Ruby-based cloud library which allows to set up a credential file to connect to many service providers and operate similarly for providing on-demand service as needed. Following figures demonstrate Fog credentials and calculation of Encryption, Load time and Trust evaluation for different Cloud Service providers (Table 2, Figs. 6, 7, 8, 9). Figure 10 provides a comparison of different approaches used for trust evaluation of each service provider. Also, Mocks created provide the in-memory simulation of

Table 2 Trust for each cloud provider

Cloud service

Trust value

Bare metal cloud

0.164625

Blue box

0.122534

Cloud sigma

0.121681

Ovirt

0.163285

VCloud director

0.173099

142

Fig. 6 Starting fog interactive

Fig. 7 Computing encryption details of cloud service

Fig. 8 Computing load time for each cloud service

B. Seth et al.

Securing Bioinformatics Cloud for Big Data …

143

Fig. 9 Computing trust for each cloud provider Fig. 10 Trust values of cloud services

services without signing up accounts each time. By running Fog on all requests in mock mode creates Excon responses for simulating real behavior.

144

B. Seth et al.

13 Conclusion and Future Outlook In this paper, we have recognized the trouble that originates in the organization of a huge quantity of rapid and heterogeneous data. A new concept recognized as Cloud Computing ought to be provided to prolong the scalability of these systems to hoard and utilize data for its explication and assimilation and ultimately for inferring knowledge to the researchers. By using clouds to store Bioinformatics big data, the companies can easily analyze massive datasets with a reduction in hardware cost and abstract the complexity, enabling proper security and scalability of resources with an increase in processing speed. However, appropriate computational atmosphere, whichever a private or public cloud must be deployed and technical expertise are mandatory for deploying the software for processing Big data in Bioinformatics using clouds. The main results of the review include a comprehensive investigation of Cloud-based Bioinformatics tools and platforms for cloud-based services including DaaS, IaaS, SaaS and PaaS solutions, Different ways how Bioinformatics and Life Sciences are using Big data and the cloud, Various real-world examples illustrating the use of Bioinformatics clouds were discussed, the paper provides a survey of the work done in this hybrid approach, and a preliminary depiction of the advantages, disadvantages, and challenges mainly security when using these tools in healthcare and bioinformatics. A trust mechanism is proposed for the users using Fog server. In our prospect work, we plan to propose an efficient Cloud data storage system using the management of Bioinformatics tools so that information is able to be stored, maintained, analyzed, and retrieved efficiently in a consistent and secure manner in the cloud.

References 1. Galbraith. Frontiers in genomic assay technologies the grand challenges in enabling dataintensive biological research. Front. Genet. (2011). https://doi.org/10.3389/fgene.2011.00026 2. Merelli, I.: Managing, analyzing, and integrating big data in medical bioinformatics: open problems and future perspectives. BioMed Res. Int. Hindawi Publishing Corporation article id:134023 (2014). http://dx.doi.org/10.1155/2014/134023 3. Oliveira, G.S.S., Edward, R., et al.: ACOsched: a scheduling algorithm in a federated cloud infrastructure for bioinformatics applications. Bioinf. Biomed. (BIBM) IEEE (2013). https:// doi.org/10.1109/bibm.2013.6732620 4. Marx, V.: The Big Challenges of Big Data, Technology Feature Nature 255, vol. 498 (2013) 5. Driscoll, O.A.: Big data, hadoop and cloud computing in genomics. J. Biomed. Inf. Elsevier (2013). http://dx.doi.org/10.1016/j/jbi.2013.07.001 6. Yang, A., Troup, M., et al.: Scalability and validation of big data bioinformatics software. Comput. Struct. Biotechnol. J. Elsevier 15, 379–386 (2017) 7. Fernandez, A., del Rio, S., et al.: Big data with cloud computing: an insight into the computing environment. In: MapReduce Programming frameworks, vol. 4, pp. 38–409. Wiley (2014) 8. Ibrar, Y., Salimah, M., et al.: The rise of big data on cloud computing. Inf. Syst. Elsevier 47, 98–115 (2014) 9. Oracle and Big Data. http://www.oracle.com/us/technologies/big-data/index.html

Securing Bioinformatics Cloud for Big Data …

145

10. Thakur, R.S., Bandopadhyay, R.: Role of cloud computing in Bioinformatics research for handling the huge biological data. In: Biology of Useful Plants and Microbes Chapter 20 Narosa Publishing House 11. Nemade P.: Big Data in bioinformatics & the era of cloud computing. IOSR J. Comput. Eng. (IOSR-JCE) 14(2), 53–56 (2013). e-ISSN: 2278-0661 12. Mu, A., Kuo, H.: Opportunities and challenges of cloud computing to improve health care services. J. Med. Internet Res. 13(3) (2011). https://doi.org/10.2196/jmir.1867 13. Luo, J.: Big data application in biomedical research and healthcare. Biomed. Inf. Insights Libertas Acad. 8 (2016). https://doi.org/10.4137/bii.s31559 14. Hua, G.J., Chuan Yi ,T., et al.: Cloud computing service framework for bioinformatics tools. BIBM IEEE (2015). https://doi.org/10.1109/bibm.2015.735899 15. Jerico, R., Bethwaite, B., et al.: Development of a cloud-based bioinformatics training platform, briefings in bioinformatics, pp. 1–8 (2016). https://doi.org/10.1093/bib/bbw032 16. Calabrese, B., Cannataro, M.: Cloud Computing in Bioinformatics: Current Solutions and Challenges (2016). http://doi.org/10.7287/peerj.preprints.2261v1 17. Shanahan, P.H.: Bioinformatics on the Cloud Computing Platform Azure 9(7) Plusone e102642 (2014) 18. Prachi, S.: Big Genomic data in bio-informatics cloud applied microbiology: open access 2(2) (2016). https://doi.org/10.4172/2471-9315.1000113 19. Hsu, H.C.: Biocloud: cloud computing for biological, genomics, and drug design. In: BioMed Research International Hindawi Publishing Corporation Article Id 909470 (2013) 20. Samuel, A.V.: CloVR: A Virtual machine for automated and portable sequence analysis from the desktop using cloud computing. BMC Bioinf. http://www.biomedcentral.com/1471-2015/ 12/356 (2015) 21. Dowlin, N., Laine, K., et al.: Manual for using homomorphic encryption for bioinformatics. Proc. IEEE 105(3) IEEE (2017) 22. Misirli, G., Madsen, C., et al.: Constructing synthetic biology workflows in the cloud. Eng. Biol. IET 1(1), 61–65 (2017). https://doi.org/10.1049/enb.2017.0001 23. Moghaddasi, H., Tabrizi, T.A.: Applications of cloud computing in health systems. Global J. Health Sci. 9(6) (2017) 24. Kumar, V.: Cloud computing using bioinformatics MapReduce applications. In: Colossal data Analysis, and Networking (CDAN), IEEE (2016). https://doi.org/10.1109/cdan.2016.7570893 25. Calabrese, B.: Cloud computing in bioinformatics: current solutions and challenges. PeerJPreprints (2016). http://doi.org/10.7287/peerj.preprints.2261v1 26. Celesti, A.: New trends in Biotechnology: the point on NGS cloud computing solutions. In: IEEE Workshop on ICT Solutions for eHealth (2016). 978-1-5090-0679-3/16/$31.00 27. Guan, X., et al.: Cancer metastases: challenges and oppurtunities. Acta Pharmaceutica Sinica. B. 5(5):402–418 (2015) 28. Afgan, E., Krampis, K., et al.: Building and provisioning bioinformatics environments on public and private clouds. MIPRO IEEE (2015). https://doi.org/10.1109/mipro.2015.7160269 29. Lukas, F., Tomislav, L., et al.: Delivering bioinformatics MapReduce applications in the cloud. MIPRO (2014). https://doi.org/10.1109/mipro.2014.68595930 30. Coutinho, R., Drummond, L., et al.: Evaluating grasp-based cloud dimensioning for comparative genomics: a practical approach. In: Cluster Computing (CLUSTER). IEEE (2014). https:// doi.org/10.1109/cluster.2014.6968789 31. Lin, L.Y.: Enabling large scale biomedical analysis in the cloud. BioMed Research International, Hindawi Publishing Corporation (2013). http://dx.doi.org/10.1155/2013/185679 32. Gabriel, D.: Food production vs biodiversity: comparing organic and conventional agriculture. J. Appl. Ecol. 50(2) (2013) 33. Che, H.L.: Cloud computing-based tagSNP selection algorithm for human genome data. Int. J. Mol. Sci. 16, 1096–1110 (2015) 34. Yixue, L., Chen, L.: Big biological data: challenges and opportunities. In: Genomics Proteomics Bioinformatics (2014). http://dx.doi.org/10.1016/j.gpb.2014.10.001

146

B. Seth et al.

35. Nguyen, T.: CloudAligner: a fast and full featured MapReduce based tool for sequence mapping. BMC Res. Notes 4(1), 171 (2011) 36. Schatz, C., Langmead, B., et al.: CloudBurst: highly sensitive read mapping with MapReduce. Bioinformatics 25(11), 1363–1369 (2009) 37. Luca, P.: MapReducing a genomic sequencing workflow. In: Proceedings of the 2nd International Workshop on MapReduce and its Applications, pp. 67–74. ACM (2011) 38. Schatz, C.M.: Cloud computing and the DNA data race. Nature Biotechnol. 28(7), 691–693 (2010) 39. Langmead, B.: Searching for SNPs with cloud computing. Genome Biol. 10(11) (2009) 40. Gunarathne: Cloud computing paradigms for pleasingly parallel biomedical applications. In: Proceedings of the 19th ACM International Symposium on High Performance Distributed computing, HDPC, pp. 460–469. ACM (2010) 41. Ben: CloudScale RNA-sequencing differential expression analysis with myrna. Genome Biol. 11, 1–11 (2010) 42. Hong, D.: RNA sequence analysis tool on the cloud. Bioinformatics 28, 721–723 (2012) 43. Jourdren, L.: Eoulsan: a cloud computing base framework facilitating high throughput sequencing analysis. Bioinformatics 28, 1542–1543 (2012) 44. Hydra, L.S.: A scalable proteomic search engine which utilizes the Hadoop distributed computing framework. BMC Bioinf. 13 (2012) 45. Matsunaga, A.: CloudBlast: combining MapReduce and virtualization on distributed resources for bioinformatics applications. In: IEEE Fourth International Conference On eScienceIndiana, pp. 222–229. USA (2008) 46. Niemenmaa, M.: Hadoop-BAM: directly manipulating next-generation sequencing data in the cloud. Bioinformatics 28, 876–877 (2012) 47. Merriman, B.: SeqWare query engine: storing and searching Sequence data in the cloud. BMC Bioinf. 11 (2010) 48. McKenna, A.: The genome analysis toolkit: MapReduce framework for analyzing nextgeneration DNA sequencing data. Genome Res. 1297–1303 (2010) 49. Huang, H.L.: BlueSNP: a package for highly scalable genomic-wide association studies using Hadoop clusters. Bioinformatics 29, 135–136 (2013) 50. Angiuoli, B.D., Matalka, M., et al.: CloVR: a virtual machine for automated and portable sequence analysis from the desktop using cloud computing. BMC Bioinf. 12 (2011) 51. Krampis, K.: Cloud BioLinux: pre-configured and on-demand bioinformatics computing for the genomics community. BMC Bioinf. 13 (2012) 52. Shi, L., Wang, Z., et al.: A case study of tuning MapReduce for efficient bioinformatics in the cloud. Parallel Comput. 61, 83–95 (2017) 53. MeteAkgun, A., Sagiroglu, S., et al.: Privacy preserving processing of genomic data: a survey. J. Biomed. Inf. Elsevier 56, 103–111 (2015) 54. Mushegian. Grand challenges in Bioinformatics and Computational Biology. Front. Genet. (2011). https://doi.org/10.3389/fgene.2011.00060 55. Heitor, H., Aleteia, P., et al.: Attribute-based access control in federated clouds: a case study in bioinformatics. Inf. Syst. Technol. IEEE (2017). https://doi.org/10.23919/cisti.2017.7975855 56. Yamamota, U., Oguchi, M.: A decentralized system of genome secret search implemented with fully homomorphic encryption. In: Smart Computing (SMARTCOMP), IEEE (2017). https:// doi.org/10.1109/smartcomp.2017.7946977 57. Namasudra, S., Roy, P., et al.: Time efficient secure DNA based access control model for cloud computing environment. FGCS Elsevier 73, 90–105 (2017) 58. Silva, S.B.S., Deborah, H.M., et al.: Secure and robust cloud computing for high-throughput forensic microsatellite sequence analysis and data basing. Forensic Sci. Int. Genet. Elsevier 31, 40–47 (2017) 59. Abdulunabi, M., Haqbi, A., et al.: A distributed framework for health information exchange using smartphone technologies. J. Biomed. Inf. Elsevier 69, 230–250 (2017) 60. Goyat, S., Jain, S.: A secure cryptographic cloud communication using DNA cryptographic technique. In: Inventive Computation Technologies (ICICT), IEEE (2016). https://doi.org/10. 1109/inventive.2016.7830158

Securing Bioinformatics Cloud for Big Data …

147

61. Nepal, S.: TruXy: trusted storage cloud for scientific workflows. IEEE Trans. Cloud Comput. 5(3), 428–442 (2016). https://doi.org/10.1109/tcc.2015.2489638 62. Siddaramappa, V., Ramesh, B.K.: Cryptography and bioinformatics techniques for secure information transmission over insecure channels. Appl. Theoret. Comput. Commun. Technol. IEEE (2015). https://doi.org/10.1109/icatcct.2015.7456870 63. Liu, B., Madduri, R., et al.: Cloud-based bioinformatics workflow platform for large-scale next-generation sequencing analyses. J. Biomed. Inf. Science direct 49, 119–133 (2014) 64. Razick: The eGenVar Data management system-cataloging and sharing sensitive data and metadata for life sciences, database (2014) 65. Rodrigues: Analysis of the security and privacy requirements of cloud-based electronic health record systems. J. Med. Internet Res. (2013) 66. Alex: Data management in the cloud: challenges and oppurtunities. Mogan & Claypool Publishers. ISBN:9781608459247 (2011) 67. Ney, P., Koscher, K., et al.: Computer security, privacy, and DNA sequencing: compromising computers with synthesized DNA, privacy leaks, and more. In: 26th UNENIX Security Symposium (2017). ISBN: 978-1-931971-40-9 68. Hamid Abdulaziz, H., Rahman Mizanur, Md.S.K., et al.: A security model for preserving the privacy of medical big data in a healthcare cloud using a fog computing facility with pairingbased cryptography, IEEE ACCESS (2017). https://doi.org/10.1109/access.2017.2757844 69. Gonzalez, N., Goya, W., et al.: Fog Computing: Data Analytics and Cloud Distributed Processing on the Network Edges, IEEE (2016). 978-1-5090-3339-3/16/$31.00_c 70. Stojmenovic, I., Wen, S.: The fog computing paradigm: scenarios and security issues. In: Proceedings of the 2014 Federated Conference on Computer Science and Information Systems, ACSIS, vol. 2, pp. 1–8. IEEE (2014). https://doi.org/10.15439/2014f503 71. Yi, S.: Security and Privacy Issues of Fog Computing, WASA (2015). https://doi.org/10.1007/ 978-3-319-21837-3_67 72. Hashem, T., Yaqoob, I., Anuar, B., et al.: The rise of “big data” on cloud computing: review and open research issues. Inform Syst 47:98–115 Elsevier (2015)

Enhancing Gadgets for Blinds Through Scale Invariant Feature Transform Raman Kumar and Uffe Kock Wiil

Abstract ICT can help blind people in movement and direction-finding tasks. This paper proposes a new methodology for safe mobility based on scale invariant feature transform (SIFT) that is expected to lead to higher precision and accuracy. Various existing gadgets for visually impaired are examined, and the conclusion is that the proposed methodology can enhance these gadgets. Keywords Scale invariant feature transform · Visually impaired · Partially-sighted and blind people

1 Introduction One of the biggest recent advances in technology is closely related to the use of mobile technology. The Internet of Things (IoT) era enables plenty of information to be extracted that can be fundamental in decision and recommendation making, such as anticipating citizen problems and providing them with better services [1, 2]. The technological evolution has led to higher processing speeds, which made new applications emerge at a faster pace. Smartphone features, such as navigation, sensoring and location-based information, opens a new world of possibilities. Designing devices to people with some kind of visually disability—visually impaired people, partially-sighted people, blind people—(we will use the term PVD—people with some visual disability) is a big challenge and subject to plenty of current and emerging research [3, 4].

R. Kumar (B) Department of Computer Science and Engineering, I K Gujral Punjab Technical University, Kapurthala, Punjab, India e-mail: [email protected]; [email protected] U. K. Wiil The Maersk Mc-Kinney Moller Institute, University of Southern Denmark, Odense, Denmark e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_9

149

150

R. Kumar and U. K. Wiil

According to the World Health Organization [5] there are 285 million visually impaired people worldwide of whom 39 million are blind. People 50 years and older are 82% of all blind. The major causes of visual impairment are uncorrected refractive errors (43%) and cataract (33%); the major cause of blindness is cataract (51%). Visual impairment is a major global health issue: the preventable causes are as high as 80% of the total global burden [5]. Provision of support for safe mobility for PVD falls under the umbrella term: assistive technology. In 1960, assistive technology was playing a key role in solving daily information transmission-based techniques. Today, it can provide impaired people with a sense of external environment using advanced algorithms [6]. Due to the magnitude and geographical distribution of visual impairment and causes, it is an important challenge to investigate new ways to support safe mobility of PVD and draw upon the existing research results from various fields. Scale Invariant Feature Transform (SIFT) algorithm assists the visual impaired and enable devices for recognition of surfaces and objects [7, 8]. SIFT features are mostly used in object recognition. These features are invariant to changes in scale, two-dimensional translation and rotation transformations. SIFT features however, are of very high dimension and large number of SIFT features are generated from any image. The large computational effort associated with matching all the SIFT features for recognition tasks, limits its application to object recognition problems [8–11]. Image matching is a fundamental aspect of many problems in computer vision, including object or scene recognition, solving for three-dimensional structure from multiple images, stereo correspondence, and motion tracking. Lucas-Kanade [12] and Random Sample Consensus (RANSAC) [13] are two algorithms that help in the major stages of computation used to generate the set of image features. These algorithms aim to use an iterative method to estimate parameters of a mathematical model from a set of observed data that contains outliers, when outliers are to be given no influence on the values of the estimates. Therefore, it can also be interpreted as an outlier detection method. The main contribution of this paper is the proposal of a feature matching methodology that uses a combination of Lucas-Kanade and RANSAC and leads to more enhanced precision and accuracy features. We use the Least Median of Square regression (LMedS) feature to overcome some of the limitations found in previous systems. The rest of this paper is organized as follows: the Sect. 2 presents a literature review of previous work that have been developed to address the issue we are dealing with. Section 3 describes the proposed methodology. Then, we present results of the proposed methodology (Sect. 4) and conclude with future lines of work (Sect. 5).

2 Previous Work With the challenges that are present in the daily lives of PVD, new technology must focus on the use in public spaces. Due to lack of precision and accuracy parameters, there is the need for additional work in the field of gadgets and assistive devices for

Enhancing Gadgets for Blinds Through Scale …

151

visually impaired. GPS is an indispensable tool for navigation for PVD, but it has been reported that navigation solutions lack precision information [12, 13]. A proposal to navigation scheme may be possible through Floyd-Warshall algorithm for suggesting shortest paths [10, 11]. In previous papers on the Lucas-Kanade and RANSAC algorithms, limited precision and accuracy aspects were discussed [14]. Bharambe et al. developed an embedded device to act as an eye substitution for the visually impaired people that helps in directions and navigation [15]. Their aim was to develop an affordable technology, which can substitute eyes for blind people. As a first step to achieve this goal they decided to make a Navigation System for the Blind. Their device consists of the following 2 parts: (1) Embedded Device: can be used to detect local obstacles such as walls/cars/etc. using 2 ultrasonic sensors to detect the obstacles and vibrator motors to give tactile feedback to the blind. (2) Android App: will give the navigation directions. The solution can be installed on any android device: cellphone/tablet/etc. [15]. An assistive device for blind people was introduced to improve mapping of the user’s location and positioning the surrounding objects using two functions that are based on a map matching approach [6]. Lowe worked on feature extraction techniques using SIFT such as at the first time Harris detector algorithm is evolved which is capable of immunity and invariance to intensity and rotation changes [7, 8]. The SIFT algorithm proposed by Lowe as described below. The cost of extracting these features is minimized by taking a cascade filtering approach, in which the more expensive operations are applied only at locations that pass an initial test. Following are the major stages of computation used to generate the set of image features: 1. Scale-space extrema detection: The first stage of computation searches over all scales and image locations. It is implemented efficiently by using a difference-ofGaussian function to identify potential interest points that are invariant to scale and orientation. 2. Keypoint localization: At each candidate location, a detailed model is fit to determine location and scale. Keypoints are selected based on measures of their stability. 3. Orientation assignment: One or more orientations are assigned to each keypoint location based on local image gradient directions. All future operations are performed on image data that has been transformed relative to the assigned orientation, scale, and location for each feature, thereby providing invariance to these transformations. 4. Keypoint descriptor: The local image gradients are measured at the selected scale in the region around each keypoint. These are transformed into a representation that allows for significant levels of local shape distortion and change in illumination. This approach was named Scale Invariant Feature Transform (SIFT), as it transforms image data into scale-invariant coordinates relative to local features.

152

R. Kumar and U. K. Wiil

Later, Zhang and Chen introduced two parallel SIFT feature extraction algorithms using general multi-core processors, as well as some techniques to optimize the performance on multi-core [16]. Kim et al. introduced an object recognition processor, which was integrated with ten processing units for task-level parallelism; and it contained a single instruction multiple data (SIMD) instruction to exploit the data-level parallelism [17]. Selviah et al. describes 3D Simulated Real Interior Environments for Mobile Devices [17]. Bonato proposed a hardware architecture of SIFT which is able to detect features up to 30 frames per second [18, 19]. Yao proposed an architecture of optimized SIFT feature detection for an FPGA implementation of an image matcher. The feature detection module took 31 ms to process a typical VGA image [20]. Kim and Lee proposed a new hardware organization to implement SIFT with less memory and hardware cost but only 553 feature points/frame [21]. Wang et al. proposes a new FPGA based embedded system architecture for feature detection and matching but the robustness to rotation and scale change of the proposed method is weak [22]. Zhong et al. presents a low-cost embedded system based on a new architecture that integrates the Wag features [23]. Huang analyzed the time consumption of each part of SIFT by running the SIFT algorithm on a 2.1 GHz Intel CPU and a soft-core of 100 MHz 32-bit NIOS II CPU [24]. D. Kim et al. proposed memory-centric network-on-chip (NoC) which only support pipelined task execution using ten processing elements [25]. In this paper, we are working on the challenges as given by the previous authors. In essence, none of the above authors have worked jointly on distance metric, Least Median of Square regression (LMedS), Lucas-Kanade and Random Sample Consensus (RANSAC) and SIFT. A combination of these approaches are expected to lead to more precision and accuracy.

3 Proposed Methodology Due to the problems with the initial matching algorithm further selection is required. The existing methods for further selection of matched points are commonly known as Least Median of Square regression (LMedS), Lucas-Kanade and Random Sample Consensus (RANSAC). Both methods are based on randomly selecting matched points and are both iterative methods. LMedS [26] calculates the median of the square of the error and seeks to minimize this. This section presents our main contribution, which consists of the proposed methodology as given below: • Step 1: Capture an Image/Video/Data as an input • Step 2: Distinct object from Image/Video/Data by using segmentation • Step 3: After identification of object apply SIFT

Enhancing Gadgets for Blinds Through Scale …

153

• Step 4: Image/Video/Data may follow the same steps till the feature matching condition • Step 5: SIFT is matched from the database, if feature matching condition fulfilled, we get data else not. The RANSAC algorithm consists of two steps. First, just enough feature point matching pairs are randomly chosen to compute model parameters. Next, this method checks the number of elements of the input feature point dataset, which are consistent with the model just chosen. RANSAC repeats the two steps within a specified threshold until it finds the maximum number of elements within a model. It then selects this model and rejects mismatches. However, the calculation speed of this algorithm is also rather slow. The two standard methods for selection of feature point matches between two adjacent overlapping images have both advantages and disadvantages. Therefore, a new algorithm [26] bearing Selection of Matched Pairs (SelMap) has been developed to perform the selection more accurately and faster and in this paper, we improve performance against that of the two standard algorithms in widespread use. The SIFT algorithm may be adopted for the various key parameters like Mahalanobis Distance. In essence, the Mahalanobis distance metric is appropriate when each dimension of image feature vector is dependent of each other and is of different importance. We have discussed the new aspects of the proposed algorithm. In the previous papers on the Lucas-Kanade [12] and RANSAC [13] algorithms, very few precision and accuracy aspects were discussed. Extensive discussions about the algorithms and techniques presented in this paper were carried out with the two authors over the past year. We may use SIFT algorithm as proposed methodology. The SIFT algorithm is mainly used for object recognition. The steps for the proposed methodology are given below: • Step 1: Build Gaussian scale space. • Step 2: Keypoint detection and localization, which checked for all scales, if yes then orientation assignment is done, else again Gaussian scale space is build. • Step 3: Keypoint descriptor creation. • Step 4: This keypoints descriptors are checked for all octaves, if true then SIFT feature vector is generated, else image is downscaled. The SIFT algorithm may be used to improve ergonomic and mechanical characteristics. It gives precise and accurate information to PVD for safety point of view. The design may be implemented in real time applications. This way multiple existing application may benefit from the proposed methodology.

154

R. Kumar and U. K. Wiil

4 Results By adopting the proposed methodology, we expect to be able to achieve the following benefits: • Determining obstacles around the user body from the ground to the head. • Affording some instructions to the user about the movement surface consists of gaps or textures. • Finding items surrounding the obstacles. • Providing information about the distance between the user and the obstacle with essential direction instructions. • Proposing notable sight locations in addition to identification instructions. • Affording information to give the ability of self-orientation and mental map of the surroundings. • Defining the route to select the best path. • Tracing the path to approximately calculate the location of the user. • Providing mobility instructions and path signs to guide the user and develop her/his brain about the environment. A few characteristics of the proposed methodology are presented below. Gradient magnitudes and orientations may be calculated using these formulae: m(x, y) 



(L(x + 1, y) − L(x − 1, y))2 + (L(x, y + 1) − L(x, y − 1))2

θ (x, y)  tan−1 ((L(x, y + 1) − L(x, y − 1))/(L(x + 1, y) − L(x − 1, y))) The magnitude and orientation are calculated for all pixels around the key point. Then, a histogram is created for this. Figure 1 illustrates binary histogram for proposed methodology. Figure 2 illustrates autocorrelation for proposed methodology. We process the magnitude and orientation based on the traces generation but storing the trace initial data in the header so that it will not create overhead to

Fig. 1 Binary histogram

Enhancing Gadgets for Blinds Through Scale …

155

Fig. 2 Autocorrelation

running processes. In this paper, we analyze binary histogram and autocorrelation with added dimensions magnitude and orientation.

5 Potential Use of the Proposed Methodology We have examined various technological developments that could help PVD. Figure 3 gives a list on different gadgets for PVD. Assisted Vision Smart Glasses yield a pair of glasses which enable someone who got very little sight to allow them to walk around unfamiliar places, to recognize obstacles, and to get a greater freedom.

Fig. 3 A list on different gadgets for VI and blind

156

R. Kumar and U. K. Wiil

AI Glasses around the world at CINVESTAV (Center for Research and Advanced Studies of the National Polytechnic Institute) in Mexico, introduced another smart glass which combines computational geometry, artificial intelligence, and ultrasound techniques, amongst other things, to create a useful aid for the visually impaired. Desarrollo are the glasses with stereo sound sensors and GPS technology attached to a tablet, which can give spoken directions and recognize denominations of currency, read signs, identify colors, and other things. It uses ultrasound, it can also detect translucent obstacles. Braille ebook reader employs thermo-hydraulic micro-actuation to activate Braille dots by infrared laser radiation via a micro-mirror scanning system. It’s easier to imagine it as a kind of wax material, which can go from solid to liquid with heat and be easily reshaped to create Braille dots. FingerReader technology for reading written text continues to improve and the good example of a new way of interacting. ARIANNA app is a great apps available for the blind and partially sighted, such as TapTapSee, which can recognize objects that you take a photo of and tell you what they are. The ARIANNA app (pAth Recognition for Indoor Assisted NavigatioN with Augmented perception) solves another difficult problem. Technology is supposed to make our and their lives easier. Too often accessibility is seen as something to tick off the list for developers, and there’s a missed opportunity to transform lives for the better. To an extent, the natural development of some technology brings unexpected benefits, like the recent story about Alex Blaszczuk, a paralyzed woman who credited Google Glass with boosting her confidence and aiding her independence. Consider how much more can be achieved when bright minds target disabilities like blindness specifically. Further we may also explore Apple Vis website is a great resource for finding apps, guides, and more content that’s relevant for the blind and visually impaired. Sonicguide, K-sonar, SmartSight, Vibe, KNFB Reader, and MusicEye are also meant for blinds but not tested and commercialized. In essence, Microsoft’s Seeing AI app, which helps blind and partially sighted people by narrating the world around them, has been released in the UK. Table 1 gives a comparison based on different gadgets for PVD and various technologies that could potentially enhance these. Our investigations show, that the proposed methodology have the potential to enhance all the discussed gadgets, whereas other technologies have the potential to enhance some of the gadgets.

6 Conclusion and Future Work This paper proposes a methodology for safe mobility based on the existing scale invariant feature transform techniques. The proposed methodology leads to higher precision and accuracy. Various existing gadgets for visually impaired are examined,

Yes

No

Sonicguide

No

No

Apple Vis

MusicEye

No

Google Glass

KNFB Reader

Yes

ARIANNA app

No

No

FingerReader

Vibe

No

Braille ebook reader

Yes

No

Desarrollo

No

No

AI Glasses

SmartSight

Yes

Assisted Vision Smart Glasses

K-sonar

[1]

Details

No

No

No

No

No

Yes

No

No

No

Yes

No

No

No

No

[2]

Table 1 A comparison on different gadgets for PVD

No

No

No

Yes

No

Yes

No

No

No

Yes

No

No

Yes

No

[3]

No

No

No

No

No

Yes

No

No

No

Yes

No

No

No

No

[4]

Yes

No

No

No

Yes

No

No

No

Yes

No

No

No

No

Yes

[7]

No

No

No

No

No

Yes

No

No

No

Yes

No

No

No

No

[27]

Yes

No

No

No

Yes

No

Yes

No

Yes

No

Yes

No

No

Yes

[9]

No

No

No

No

No

Yes

No

No

No

Yes

No

No

No

No

[10]

No

No

No

No

No

Yes

No

No

No

Yes

No

No

No

No

[11]

Yes

No

No

No

Yes

No

No

No

Yes

No

No

No

No

Yes

[8]

No

No

No

No

No

Yes

No

No

No

Yes

No

No

No

No

[28]

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Proposed

Enhancing Gadgets for Blinds Through Scale … 157

158

R. Kumar and U. K. Wiil

and it is concluded that the proposed methodology can be used to enhance all these existing gadgets. In the future, it is our objective that the proposed methodology is adopted as an optimized algorithm for various assistive technologies. As part of this, existing gadgets may be equipped with new functionality that allow users to choose reference points with more accuracy and precision. Acknowledgements The authors also wish to thank the anonymous reviewers for their suggestions to improve this paper.

References 1. Ifukube, T., Sasaki, T., Peng, C.: A blind mobility aid modeled after echolocation of bats. IEEE Trans. Biomed. Eng. 38(5), 461–465 (2002) 2. Gerard, L., Kenneth, M., Dawson, H.: The application of robotics to a mobility aid for the elderly blind. Robot. Auton. Syst. 23(4), 245–252 (1998) 3. Kay, L.: An Ultrasonic sensing probe as a mobility aid for the blind. Ultrasonics 2(2), 53–59 (1964) 4. Razali, M.F., Toha, S.F., Abidin, Z.Z.: Intelligent path guidance robot for visually impaired assistance. Procedia Comput. Sci. 76, 330–335 (2015) 5. Limburg, H., Kumar, R., Indrayan, A., et al.: Rapid assessment of prevalence of cataract blindness at district level. Int. J. Epidemiol. 26, 1049–1054 (1997) 6. White, C.E., Bernstein, D., Kornhauser, A.L.: Some map matching algorithms for personal navigation assistants. Trans. Res. C Emerg. Tech. 8, 91–108 (2000). https://doi.org/10.1016/ S0968-090X(00)00026-7 7. Lowe, David G.: Distinctive image features from scale-invariant keypoints. Int. J. Comput. Vis. 60(2), 91–110 (2004) 8. Lowe, D.G.: Object recognition from local scale-invariant features. In: International Conference on Computer Vision, Corfu, Greece, pp. 1150–1157, September 1999 9. Mikolajczyk, K., Schmid, C.: An affine invariant interest point detector. In: ECCV, pp. 128–142 (2002); Schaffalitzky, F., Zisserman, A.: Multi-view matching for unordered image sets, or “how do i organize my holiday snaps?” In: ECCV, pp. 414–431 (2002) 10. Floyd, R.W.: Algorithm 97: shortest path. Commun. ACM 5(6), 345 (1962). https://doi.org/ 10.1145/367766.368168 11. Warshall, S.: A theorem on Boolean matrices. J. ACM 9(1), 11–12 (1962). https://doi.org/10. 1145/321105.321107 12. Prudhvi, B.R., Bagani, R.: Silicon eyes: GPS-GSM based navigation assistant for visually impaired using capacitive touch braille keypad and smart SMS facility. In: Proceedings of the 2013 World Congress on Computer and Information Technology (WCCIT), Sousse, Tunisia, 22–24 June 2013 13. Black, A.W., Lenzo, K.A.: Flite: a small fast run-time synthesis engine. In: Proceedings of the ITRW on Speech Synthesis; Perthshire, Scotland. 29 August–1 September 2001 14. Chum, O., Matas, J.: Optimal randomized. Ransac Ieee Trans. Pattern Anal. Mach. Intell. 30(8) (2008) 15. Bharambe, S., Thakker, R., Patil, H., Bhurchandi, K.M.: Substitute eyes for blind with navigator using android. In: Proceedings of the India Educators Conference (TIIEC), Bangalore, India, 4–6 April 2013, pp. 38–43 16. Q. Zhang and Y. R. Chen, SIFT implementation and optimization for multi-core systems. In: Proceedings of IEEE International Symposium on Parallel and Distributed, pp. 1–8 (2008)

Enhancing Gadgets for Blinds Through Scale …

159

17. Selviah, D.R., Baghsiahi, H., Hindmarch, J.: 3D simulated real interior environments for mobile devices (2013) 18. Bonato, V., Marques, E., Constantinides, G.A.: A parallel hardware architecture for image feature detection. In: Woods, R., Compton, K., Bouganis, C., Diniz, P.C. (eds.) Reconfigurable Computing: Architectures, Tools and Applications. ARC 2008. Lecture Notes in Computer Science, vol. 4943. Springer, Berlin, Heidelberg (2008) 19. Bonato, V., Holanda, J.A., Marques, E.: An embedded multi-camera system for simultaneous localization and mapping. In: Bertels, K., Cardoso, J.M.P., Vassiliadis, S. (eds.) ARC 2006. LNCS, vol. 3985, pp. 109–114. Springer, Heidelberg (2006) 20. Yao, L.F.: An architecture of optimised SIFT feature detection for an FPGA implementation of an image matcher. In: Proceedings of International Conference on Field-Programmable Technology, pp. 30–37 (2009) 21. Kim, S., Lee, H. J.: A novel hardware design for SIFT generation with reduced memory requirement. J. Semicond. Technol. Sci. 13(2), 157–169 (2013) 22. Wang, J., Sheng, Z., Yan, L., Cao, Z.: An embedded system-on-a-chip architecture for real-time visual detection and matching. IEEE Trans. Video Technol. (2013). Accepted for publication 23. Zhong, S., Wang, J., Yan, L., Kang, L., Cao, Z.: A real-time embedded architecture for SIFT. J. Syst. Architect. 59(1), 16–29 (2013) 24. Huang, Ch.: High-performance SIFT hardware accelerator for real-time image feature extraction. IEEE Trans. Circuits Syst. Video Technol. 22, 340–351 (2012) 25. Kim, D.., Kim, K., Kim, J.Y., Lee, S. Lee, S.J., Yoo, H.J.: 81.6 GOPS object recognition processor based on a memory-centric NoC. IEEE Trans. Very Large Scale Integr. (VLSI) Syst. 17, 370–383 (2009) 26. Li, Z., Selviah, D.R.: Comparison of image alignment algorithms (2011) 27. Siggelkow, S.: Feature histograms for content-based image retrieval. Ph.D. thesis, AlbertLudwigs-University Freiburg (2002, December) 28. Grabner, M., Grabner, H., Bischof, H.: Fast approximated SIFT. In: Proceedings of ACCV 2006, Hyderabad, India (2006)

Part II

Applications of Computational Intelligence

A Study to Renovate Image Data Using Data Analytics Methodologies A. Prema

Abstract In the present scenario storing voluminous data becomes a challenging issue. This research work is an initiative to convert the already existing image analysis data of a pipeline study for future use. Image data is obtained by the process of digitized data concept with the application of k-means cluster algorithm and morphological operator. Data obtained through this process is voluminous and hence a new technique should be adapted to store and retrieve the data for any further use and reference. As data analytics is emerging as one of the methods to analyse data and store, this study helps to deal with the same. Data collected from oil pipeline image processing with pattern recognition which used the technique of unsupervised learning is to be converted under the process of Machine learning. The key step for this process is to first determine the data set for evaluation and then to later convert these date to the required base set, this helps to save the data in large volumes for future use. To perform this task a key study on Data Analytics and Machine Learning are required. Data analytics refers to the quality and quantity of data extracted and categorized from various resources, these data are collected to analyze the structural behaviour and pattern, which differs accordingly to the need of the organizations. The process of machine learning starts with the collection and scrutiny of data that can be obtained through direct or indirect evaluation of data, which will give a pattern for the defined image for further process. The proposed machine learning algorithm has to be fixed to study the data that is been used in the pipeline image conversion task. Structural element concept based on mathematical morphological operator helps to define the size of the data to be stored. Algorithms based on structuring element paves way for redefining the concepts to shrink and save the data for further use. This study will enhance the users to save the data from pattern recognition to Machine learning using data analytics techniques. The overall concept is to save the existing image data in data analytics techniques for future use and references. This research paper spotlights a way for an energy firm to congregate, identify and sway the distinct statistics obtained through image analysis and processing. It also helps to build a strong and A. Prema (B) Department of Computer Science, School of Computing Science, Vels Institute of Science, Technology and Advanced Studies (VISTAS), Pallavarm, Chennai, India e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_10

163

164

A. Prema

confident efficient productivity thus giving an opportunity for best practices of data management breaking down the question of data security and enhancement. This helps to avoid data theft by securing the data obtained through the image analysis and processing concepts.

Preface Renovation of everything is good to the development and it shows that the process is moving to next level under the pretext of development. In this research work style of storing data is renovated from a small level into bigger status, which paved the way for this article. When i started it was very basic but later it lead to the innovation of IDAP from BIDAP that takes care of the entire style of converting image data from bytes to pegabytes. My previous work all concentrated only in converting the image captured from bits to bytes, but the memory was not enough. So to enhance the storage capabilities the work was routed to data analytics, which uses the method of transforming small data storage into huge volume. This style of conversion will help the beginners to work more into data conversion, depending on their process and requirement base. Future work on data for researchers will aid in the development of software as well as hardware. I thank all for giving me this opportunity to explore more on data storage.

1 Introduction Pipeline image processing helps in identifying the cracks inside a pipeline. The overall functioning is explained here in the diagram. A friction wheel is placed inside an oil tank to help in the movement of digital camera or mobile robots to detect the cracks or holes present inside a pipeline. Friction is defined as a force that works against movements of objects, friction helps us to maintain stability and prevents slipping of objects that is inserted inside a pipeline. Thus a friction wheel is used to move the objects inside the oil tank. Through the video sensor that is connected to the system tracks the flow of oil inside the pipeline. A pump is used to maintain the pressure flow inside the pipeline. Computer helps to retrieve the image and using the image processing methods holes and cracks are detected, where blur images [1] are segregated restudied and investigated for clear image for further detection, this helps in detecting the flaws in the path of a pipeline (Fig. 1). Image processing plays a vital role in today’s life, because of the large abundance of data available in our day to life. Usage of image processing techniques paved way for the development of many processes. One such technique is Intelligent Character Recognition (ICR) which is automated to extract data from handwritten images or digital images in the form of jpg/png/tif/mpeg format. Input images are taken as scanned format and recognition tool does this job by applying pre-processing techniques to extract characters from the images obtained. These inputs are trained using

A Study to Renovate Image Data Using Data Analytics Methodologies

165

Friction wheel

Blur image

Computer Video Sensor

Wear Particles Oil Flow Path

Pump

Oil Tank Fig. 1 Diagramatic representation of oil pipeline image processing INPUT IMAGE

IMAGE PROCESSING

IMAGE SEGMENTATION

DATABASE FORMATTED USING DATA ANALYTICS

IMAGE RECOGNITION

DATA ENTERED IN TO DATABASE

Fig. 2 Image processing transaction using data analytics

CNN’s (Convolution Neural Network) models for identification and recognition and then write the data to MYSQL database. Text can be directly entered into the database using ICR technique by post classifying all the segments in the image set after doing proper character recognition.ICR operates by capturing image text from files and converting them into text searchable files by giving the users the ability to search through the files with text strings and capture information from them by using the copy/paste function. This technology avoids multiple entries by multiple users and thus eliminates human errors (Fig. 2).

2 Proposed Method In the entire technique image processing is the initial step to start with the process, which implements segmentation methods and extracts the required fields from the input for recognition.

166

A. Prema

2.1 Image Processing Input is obtained through image reading using a digital camera with a god resolution of 300 dpi for maximum accuracy. If the image is not clear evaluation is not possible with accurate values. Pre-processing of the image [2] is the main process, because it involves the procedure of removal of noise from any image. It aims to improve the image quality that overwhelms unwanted biases or improves some image features for further processing.

2.2 Image Segmentation Image segmentation is the process of dividing an image [3] into multiple parts. This is typically used to identify objects or other relevant information in digital images. In this section images from the digital camera is segments to find out the defects in the pipeline path.

2.3 Image Recognition The process of converting digital image using electronic means into digital data is image recognition, where it helps to identify the holes and cracks of a pipeline from the image captured.

2.4 Data Entered into Database The digital data from image recognition is stored in a database to read the 0s and 1s from the image for further mathematical process to identify the exact position of the pipeline cracks or holes [4].

3 Data Gathered from an Oil Pipeline The main process of information gathering and making it digitized is to reduce the risk of accidents that occur during oil transportation. Pipelines are now made of fibres alongside oil and gas line stretch to transmit real time data to operate in an efficient way. These fibre based helps to gather more data and the movement of an object inside the pipeline will also reduce the risk of accidents and other hazards as the oil

A Study to Renovate Image Data Using Data Analytics Methodologies

167

flow is watched live. Thus the data collected from this method is then processed and saved for future references.

3.1 Database Formatted Using Data Analytics The data in the database is stored temporarily and cannot be used for future reference. To overcome this problem data analytics methodology is implemented to preserve the data for future references. The processed data is stored for future comparisons for a pipeline image study. Data Analytics refers to qualitative and quantitative techniques and processes used to enhance productivity and business gain. Data is extracted and categorized to identify, analyze behavioural [5] data, patterns and techniques that vary according to organizational needs. The process of examining data as sets in order to draw conclusions about the information they contain, with the aid of specialized systems and software. This technique is widely used for commercial industry to enable organizations to make more-informed business decisions and also by scientists, researchers to verify or refute scientific models, theories and hypotheses. (Tech Target)

3.2 Data Analytics in Oil Pipeline Innovations in Oil and Gas industry technology have the potential to impact everyone. To access a safe, clean and for cheap extraction technology has made a very big business opportunity. The innovation in digital world has lead to greater advancements in the big data acquisition and analysis on a large scale. The data retrieved can also be evaluated online for space constraints but for future reference it must be saved in a huge manner. Basically there are few inadequacy which is overcome using data analytics in a competent way. To name a few: • Cost optimization in the image data processing • Exploration on data obtained • Storing possibilities.

4 Renovation of Image Data Renovation of image data depends on the classification model, tedious or simple decision depends upon the type of process image data is going to undertake. Simple models are easy but when it comes to the identification of more cracked image then algorithms that follow deep learning like CNN (Convolution Neural Networks) [6]

168 Fig. 3 IDAP architecture

A. Prema USER INTERFACE

IMAGE SYNTHESIS

ARTIFICIAL INTELLIGENCE MODELLING LANGUAGE(AIML)

PATTERN MATCHING

DATABASE DICTIONARY

should be followed for image recognition with dense proposals. After the conversion of image into required data it is stored for future use. Image Data Analytics Platform (IDAP) is our main platform to bring the latest image analysis tool to end-users. Large complex operations that are measured in terms of petabytes are used here by using very simple and minimal commands. With the help of the commands, images are searched based on the available data instead of the Meta data [7] as it is a real time exploration inside the pipeline for crack identification. The requests are performed with high speed and efficiency is achieved at the best in this process.

4.1 Technology Involved IDAP is developed based on the latest big data framework (Apache Spark). It imparts a measurable, flaw lenient and distributed backend to handle large datasets (giga, tera and peta bytes). This technology involves image analysis on numerous images by implementing simple query language 4QL, it is an extension of SQL that offers an imaging functionality that doesn’t require a developer experience (Fig. 3).

5 User Interface User interface is a method by which the user and a computer system interact, with the help of input devices and software. In our process a computer is connected to the digital camera or mobile robot accordingly, to get the images [8] from pipeline path.

A Study to Renovate Image Data Using Data Analytics Methodologies

169

5.1 Image Synthesis Image synthesis is the processes of creating new image from some form of image description. Images that are typically synthesized are: • Test Patterns: scenes with simple 2D geometric shapes • Image notes: image containing random pixel values, usually generated from specific parameterised [9] distributions • Computer Graphics: Scenes or images based on geometric shape descriptions, often the models are 3D, but may also be 2D [10]. Synthetic images are often used to verify the correctness of operators by applying them to known images. They are also often used for teaching purposes, as the operator output on such image is generally “Clean and clear”, whereas noise and uncontrollable pixel distributions in real images make it harder to demonstrate unambiguous results. The image should be binary, grey level or colour. The image obtained from the output process is synthesized for further processing. It identifies the colour image captured and sends for further processing.

5.2 Artificial Intelligence Markup Language Artificial intelligence modelling language is an XML based markup language used to create artificial intelligent applications. It makes it possible to create human interfaces while keeping the implementation simple to program, easy to understand and highly maintainable. AIML is used as an interface [11] between the user and the system to get the images from the pipeline pathway and later send it for further processing.

5.3 Pattern Matching Pattern matching is defined as the technique used for checking and locating specific sequences of data among raw data or a sequence of tokens. Unlike pattern recognition, the match has to be exact in the case of pattern matching. Pattern matching is one of the most fundamental and important paradigms in several programming languages. Many applications make use of pattern matching as a major part of their tasks. Image data obtained is matched for already existing data to make the job easy, which will lead to rapid identification of the data with error [12].

170

A. Prema

5.4 Database Dictionary A data dictionary is a file or a set of files that contains a database metadata. It contains records about other objects in the database, such as data ownership to other objects and data. This is very important in our research work as it is going to store [13] the image for future use and this will help in comparing the new data with the existing ones. This method of saving the values obtained from a identification process is reliable and secure.

6 Conclusion Data conversion using data analytics has helped to preserve the data in a better way. As we move to digital camera on mobile robots, Scientists from Korea have developed an inch worm robot (Hygrobots) that works without an engine or electricity to implement motion inside an oil pipeline. These robots run on environmental humidity, absorbing the moisture inside the pipeline path. A mathematical model of motion is also structured for the model motion to maximize the locomotion speed. These robots are going to get better. With the help of these types of tools image recognition is going to experience a new era that will enhance the oil business. Profit will be more if the concern is going to merge the existing technique with latest technologies.

References 1. Zhang, X., Chen, S., Fan, J., Wei, X.: A grid environment based satellite images processing. In: Proceedings of the 1st International Conference on Information Science and Engineering (ICISE), 26–28 Dec 2009 2. Satyanarayan Rao, K.K. Pattanaik, M.B.: Toolkit for grid-enabled high resolution image processing. In: 2nd International Conference on Signal Processing Systems (ICSPS) (2010) 3. Prema, Dr. A., et al.: Implementing mathematical morphology techniques to detect cracks in underground pipeline images automatically. Int. J. Sci. Basic Appl. Res. (IJSBAR) 13(1). ISSN 2307–4531 (Print & Online) (2014) 4. Muresan, O., Gorgan, D., Cristea, V.: Satellite image processing applications in MedioGRID. In: Fifth International Symposium on Parallel and Distributed Computing (ISPDC ‘06), pp. 253–262 (2006) 5. Huang, H.K., et al.: Data Grid for Large-Scale Medical Image (2005) 6. Archive and analysis, multimedia. In: Proceedings of the 13th Annual ACM International Conference on Multimedia, pp. 1005–1013 (2005) 7. Germainet, C., et al.: Grid-enabling medical image analysis. In: Proceedings of the IEEE International Symposium on Cluster Computing and the Grid, pp. 487, 495 (2005) 8. Glatardet, T., et al.: Grid-enabled workflows for data intensive medical applications. In: Proceedings of the 18th IEEE Symposium on Computer-Based Medical Systems (2005) 9. Belkheir, K., Zaoui, L.: Content-based image retrieval in grid’s environment. In: Proceedings of the Fourth International Conference on Neural, Parallel & Scientific Computations, 11–14 Aug 2010

A Study to Renovate Image Data Using Data Analytics Methodologies

171

10. Montagnatet, J., et al.: Medical images simulation, storage, and processing on the European DataGridtestbed. J. Grid Comput. 2, 387–400 (2004) 11. Chiang, T.-W., et al.: A hierarchical grid-based indexing method for content-based image retrieval. In: Proceedings of the Third International Conference on International Information Hiding and Multimedia Signal Processing (IIH-MSP), vol. 01 (2007) 12. Town, C., Harrison, K.: Large-scale grid computing for content-based image retrieval. Aslib Proc. 62(4/5), 438–446 (2010) 13. Montagnat, J.: Texture-based medical image indexing and retrieval on grids, medical imaging technology, vol. 25, No. 5, Nov 2007

The Use of Modern Technology in Smart Waste Management and Recycling: Artificial Intelligence and Machine Learning Praveen Kumar Gupta, Vidhya Shree, Lingayya Hiremath and Sindhu Rajendran Abstract Waste management is one of the primary problem that the world faces irrespective of the case of developed or developing country. The key issue in the waste management is that the garbage bin at public places gets overflowed well in advance before the commencement of the next cleaning process. It in turn leads to various hazards such as bad odor and ugliness to that place which may be the root cause for spread of various diseases. The increase in population, has led to tremendous degradation in the state of affairs of hygiene with respect to waste management system. The spillover of waste in civic areas generates the polluted condition in the neighboring areas. For eliminating or mitigating the garbage’s and maintains the cleanness, it requires smartness based waste management system. The need of proper waste management does not end with just collection and proper dispose of garbage. It continues to the level of landfills and the amount that we can possibly recycle. Recycling is estimated to be highly useful given that our dependency on raw products reduces, besides the reduction of waste and subsequent landfills. Once the recycling is done to sort metals, plastics, and glass articles, the use of biodegradable waste can be extended beyond fertilizers and manure. The metals can be reused and the plastics can be diverted from the landfills, which otherwise leads to choking of the earth. The glass materials can be broken and melted back to form new articles after deep cleaning. This chapter aims to understand the use of machine learning and artificial intelligence in the most potential areas and the ultimate need to completely replace the human interaction.

P. K. Gupta (B) · L. Hiremath Department of Biotechnology, R. V. College of Engineering, Bangalore 560059, India e-mail: [email protected] V. Shree Department of Electronics and Instrumentation, R. V. College of Engineering, Bangalore 560059, India S. Rajendran Department of Electronics and Communication, R. V. College of Engineering, Bangalore 560059, India © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_11

173

174

P. K. Gupta et al.

Keywords Location intelligence · Smart city · Graph theory optimization · Internet of things · Waste management · Automation · Artificial intelligence in recycling · Graph theory · Graph optimization · Machine learning · Artificial neural networks (ANN)

1 Introduction In one of the recent surveys conducted by World Bank, about 1.3 billion of tons of waste is generated each year. It is estimated that the numbers would rise to nearly 2.2 billion by 2025. These numbers have a significant impact on our societies and on the ecology on the whole. The more the waste we generate, the more we waste the resources and initiate hazardous situations for bio life. Waste management is one of the primary problem that the world faces irrespective of the case of developed or developing country. Improper waste management has been proven to be the sole reason for the hazardous air pollution leading to serious health issues like COPD (Chronic Obstructive Pulmonary Disease) and Asthma, for people who have their livelihood close to these Garbage bins. The failure to remove this accumulated garbage is the sole reason for breeding of mosquitoes and houseflies, which is the root cause of fatal diseases like Dengue, Malaria, etc. Nearly 235 million people are victims of breathing illness due inhalation of air with foul smell [1]. While reducing the level of human interaction in the waste management is equally essential, we need to look for industrial automation using smart waste management techniques. This is needed right from proper waste disposal, waste collection, prompt checking to prevent over flow to safely disposing them. Recycling is yet another major innovative step towards reduction of waste and to get back from the used products. Recycling industry is now dominated by dumpster robots which makes the task of sorting and recycling very easy and brings down the human intervention and exposure to waste. Through this chapter we hope to establish the information based on the latest technology which has been proposed and practically implemented with optimization as the future developments.

2 Smart Waste Management Using Artificial Intelligence In order to increase the efficiency of cleaning processes, they are several methods based on artificial intelligence and IoT, which effectively designed and implemented. The aim of an integration of ‘Smart Cities’ is the ultimate goal of each developed and developing country. With the integration of IoT and Data access networks, combinatorial optimization, Geographic Information Systems (GIS), implemented through electronic engineering, the waste management of any city can be organized easily [2]. The IoT integrated trash cans can sense and send the data about the trash volume over to the servers through the internet. Using the graph theory optimization

The Use of Modern Technology in Smart Waste Management …

175

techniques and associated algorithms, the data is processed to develop dynamic and efficient waste management. The proposed system is simulated in a realistic city of Copenhagen, using Open Data. This is pitched into the reality to attract third party investments into the benefits attained by this model and result in innovative solutions for smart cities. The established system depicts the involvement of Geographic Information Systems (GIS), applied graph theory optimization, and some concepts of machine learning. The use of sensors is essential for sending the trash volume data to the servers, which is processed and used to track and initiate the cleaning process accordingly on a routine. This ensures that the trash is cleared well before it overflows and also, prevents the unnecessary trash pickup when the trash cans are not filled. Open Data is known to be the initiator of Big Data analytics which are enhanced with cyber systems [3].

2.1 System Function The system is completely based on the indication of waste levels in the trash can by the embedded sensors. The data about the volume is transmitted to the servers over the internet. The collected data after processing is used to optimize and direct the municipality cleaning process. With this automated process, the cleaning team has new routes each day, to visit and clear trash bins based on the need and also ensure that there is no trash left unattended. The end results are a cleaner society with reduced costs and manual efforts. The collected data is used to statistically analyze the rate of filling up of these trash cans (Fig. 1).

Fig. 1 The proposed system in the smart city: Copenhagen [2]

176

P. K. Gupta et al.

2.2 System Components 1. Sensors: The amount of waste is determined by the level of waste collected from the top of the lid. The SONAR technique is used to sense the distance, using Ultrasonic Ranging Module (HC-SR04). With an accuracy of 3 mm, this technique can clearly indicate the distance in the range from 2 to 400 cm. The temperature sensors, and weight sensors along with metal detectors are useful for sorting the biodegradable, recyclable and reusable items from the trash cans [2]. Working of Ultrasonic Ranging Module: The Basic principle involved in the non-contact measurement is: 1. The IO trigger is used for a minimum of 10 μs of the high level signal. 2. The unit is designed to transmit eight 40 kHz of the triggering signal which is detected once it is received back. 3. The time lapse in the event of emission and retrieval of the signal is used to calculated and accordingly calibrated to estimate the distance of the level of the trash filled (Fig. 2). The angle between the sensors along with the distance from each other is important to overcome the ‘crosstalk’. The target level is to be mounted at a perpendicular angle from the base of the sensor [1]. 2. Microcontroller: Arduino Uno, based on ATMega328, is used as the decision making microcontroller in this prototype. With an operating voltage at 5 V and optimized power consumption of about 40–50 mA, it is sufficient to collect data and transmit it over the internet. The microcontroller has been chosen based on the parameters like minimal power consumption, memory, processing and the economic cost. 3. Access Network Interface: using the WiFi technology, the information from the trash cans can be sent to the servers through internet. One of the methods have

Fig. 2 The working of the ultrasonic ranging module [1]

The Use of Modern Technology in Smart Waste Management …

4. 5. 6.

7.

177

implemented the CC3000 Shield with the on-board antenna as it is known to have exceptional coupling with the microcontroller chosen. Battery output: Since the working of this system is based on a daily report generation, the battery is chosen to facilitate a long duration of use. Database: The data collected from the trash cans are organized into the data systems, using the software such as MySQL. Artificial Intelligence: from the data collected, the statistical analysis helps in the determination of the possible rates of filling the trash cans and the appropriate routes to be followed regularly by the cleaning team. Also, the timings of the data collected from trash cans emphasizes on the locations to be given priority during the different times of the day. Some places might see a regular filling of trash cans in the mornings while others in the late nights. Also, there are some locations like the schools, offices, houses which need immediate cleaning compared to open fields with lesser human interaction. So once the artificial intelligence is boosted to the system, the routes are defined and optimized based on the timings and the corresponding need [4]. Optimization techniques: with the recorded data, the system uses the artificial intelligence and optimization techniques [5] to reduce the transportation expenditure of costs and time, by directing the best routes to empty the filled trash cans. This is optimized keeping in mind the traffic, and the driving distance and time. The end of the optimization results in a highly efficient route to be followed by the truck drivers, enhanced with the GPS location services [6].

2.3 The Smart Dustbin The cylindrical structure is associated with a piston that is useful for compression of the garbage. The trash can plate is attached to the cylinder and the leaf switch is to be suspended upside down [7] through the side hole. The leaf switch level is placed at a point lower than the maximum level. This is essential for the precautionary measures of the garbage overflow, in cases of fault from the cleaning team. The compressing plate can reach down to press the switch. Once the threshold level is reached, the garbage is prevented to be dumped inside the trash can in order to avoid overflow (Fig. 3).

2.4 The Optimization Techniques and the Corresponding Algorithms Opted 1. Shortest path Spanning Tree (SPST): The SPST technique is used to find the shortest distance between two trash cans in the pre entered graph data of the streets and lanes in the outlay of the Smart city. The streets are associated with

178

P. K. Gupta et al.

Fig. 3 The design of the smart dustbin [1]

the edges and the joining points are considered as the vertices. This is essential to optimize the routes for trash pickup from the data received from all the trash cans [8]. 2. Genetic Algorithms (GA): The optimization of the trash collecting cycles is a combinatorial problem, which is to be duly similar to the ‘The minimum Linear Arrangement Problem’ [9] and the popular ‘The Traveling Salesman Problem’ [10]. The GA techniques are quite efficient when it comes to highly optimized driving routes for the trash clearance [11]. 3. K-Means: another NP-Hard problem, it is especially complex when it comes to solving hard clustering problems. K-means is highly efficient method to be used as solution in situations of huge complex cluster parameters.

2.5 The Evidence of Practical Output of the Proposed Model With our simulated city of Copenhagen, Denmark, the working was tested with the real data from GIS of the streets and trash can locations: 3046 trashcans with about 18 division of trash cleaning teams. The dots represent the location of the trash cans and the different colors are used to indicate the different division teams, in charge of those trash bins (Fig. 4). The trashcans were modelled based on the filling in the Poisson’s distribution. The individual trash can i, is assigned as per the assigned time to reach Fi, with a stipulated λ  7 for 7 regular days of the week. Also, the trashcan i, can be assigned a daily filling volume of about DFi, in accordance to the poisons distribution with

The Use of Modern Technology in Smart Waste Management …

179

Fig. 4 The street network and the location of the trashcans, under study in Copenhagen, Denmark [2]

λ  1/Fi, where d corresponds to each day of the period under test. This helps us determine the variance across the trash cans being filled. The route creation for the division teams is associated with the optimization techniques and algorithms adopted. For each individual team, a particular pint was chosen to be set as the headquarters and all the routes assigned began from this point and ended right here. The economic restrictions were based on the collection costs, for which various parameters like the number of workers, the driving speed, driving time, the fuel consumption and the internet requirement sand one time set up costs are included and analyzed against the efficiency of cleaning. The performance parameters are evaluated based on different scenarios which optimize the expenditure and also help the ultimate goal of a clean Smart city [2].

2.6 The Alternative Technology for Data Transfer The other alternative technology used in the information transmission is using the RFID tags and RFID readers. It is based on the preliminary radio technology. The RFID tag is the combination of the antenna and the embedded microchip which store an identical serial number to track the object or the person. With the RFID tag, the radio frequency is transmitted over through the Reader’s Antenna. The RFID tag modulated signals is received by the reader. The RFID tag is responsible to store the trash data and transmit it to the RFID reader [12]. The active tags have longer range of service with their own internal battery to support their power consumption. The Passive tags depend on the readers to suffice the operating power and therefore have a lower range of just a few meters [13]. The RFID has the radio frequency transmitted once the reader in switched ON. Once the RFID tag comes near to the reader antenna, the data is received and reverted

180

P. K. Gupta et al.

back to the RFID reader. The most common reader used in the EM-18 and it is a passive tag with the tag ID shifted to Arduino Microcontroller. The Working of the RFID system: The microchip contains a unique ID that integrates the circuitry for the proper functioning of the tag. The internal EEPROM is essential to trace and keep the unique ID. The tag used is passive so the antenna receives the power and the radio frequency signals from the RFID reader. To recognize the RFID tag, the reader sends the data back to the tag. The coil receives the signals in the form of alternating current and passes it to the microchip [13] (Fig. 5). The RFID tags and the readers are to be in the same frequency for the communication. The wide range of frequencies is left as the option for optimization and as per the requirements. The low frequency is around 125 kHz and the high frequency is about 13.56 MHz with the ultra-high frequencies varying between the 860 and 960 MHz. Most of the applications are also based on the 2.45 GHz frequency too. The RFID tags are usually of CMOS IC EM4102, which is a passive RFID tag. The reader generates a frequency of about 125 kHz and in occurrence of the 125 kHz passive RFID tag, it will get energized from it. The entire system is programmed to send a notification to the server once the trash cans reach a certain level. The RFID reader is interfaced with the microcontroller to enhance the process of verification. As the RFID tag interrupts the RFID reader, the ultrasonic sensor calculates the level of trash filled in the dustbin. Based on the

Fig. 5 The working of the RFID tag [1]

The Use of Modern Technology in Smart Waste Management …

181

level indicated the cleaning process is initiated. Usually the RFID tag is placed in the trash bin and the RFID reader is placed with the antenna in the truck.

2.7 The Conclusion of Smart Waste Management The efficient and proactive cleaning of the trashcans with the optimized technology to save the manual effort, time and cost is highly appreciated and is in practice in most societies. The model establishes a stronger waste collection system which reduces the overflowing trash cans by the factor of 4. As the regular trash pickup is indicated by the trash cans, the regularity of the check improves, thereby increasing the costs by a margin of 13–25%. However, when the results are tallied across the required goal of clean environments with regularized cleaning, the objective is achieved with a high success rate.

3 Sorting and Recycling of Waste The proper trash collection does not end the problem of waste management. The trash collection just ensures that the garbage does not affect the livelihood of people and animals. But piles of trash are filling up the land mines and ending up in the oceans. This is degrading the quality of life and the ecosystem. In order to efficiently dispose the garbage, it is necessary to sort them into biodegradable, recyclable and nonbiodegradable substances. This ensures reutilization of the biodegradable substances to enrich the soil and aggravate cultivation. The recycling of waste materials right from the metals to the paper and plastics, is high on trend. Since manual sorting is difficult and unhygienic, there are several methods which involve robots and concepts of machine learning. This section deals with the techniques involved by the latest technologies to help recycle the massive piled up unsegregated trash all over.

3.1 Introduction The recycling of materials like the steel, glass, and plastics helps to reduce the dependency and extraction of raw materials along with preventing the pile up of dump yards. The latest research has resulted in the implementation of machine learning in the strategies to be used for material segregation, using an induction algorithm (AIMS). This section highlights the procedure detecting plastic, glass and metal objects from huge masses of trash by using light transmission techniques. The realization of machine learning in practical scenario with high intelligence automation used in garbage sorting and recycling is the future of robotics. A general municipal cleanup is unnoticed until the cost of disposal of the waste through landfills or other

182

P. K. Gupta et al.

means increase. The revenue potential of recycling can be boosted with established intelligence automation methods.

3.2 Machine Learning: A Boon to Automation While machine learning is aiming at the ultimate goal of zero human involvement, it is used in many ways to simplify lifestyle of all. The unforeseen and less debated part of machine learning resorts to the problems of unemployment and the corresponding financial crisis that the world needs to prepare for. Though machine learning is doing exceptionally well with respect to the human errors, we need to focus on developing technology in areas which needs to be away from human intervention. This could be places like nuclear power plants, space shuttles, aircraft and turbine testing, and so on. While we integrate automation with the recycling industry, the artificial intelligence used in the background for these super sorting bots is getting smarter by the day. The future of human civilization is exaggerated in science fiction films using teleportation, time machines, giant holograms, flying cars and movable apartments. But the subject of how to handle the waste that is slowly killing the ecosystem and indirectly us, is often left without a thought. The Materials Recycling Facilities (MRF) is the new innovative solution to it all.

3.3 Statistics In the era of self-absorbed millennials, the landfills are the major cause of increasing Greenhouse gas (GHG) emissions and the significant factor of drastic global climate change. The ‘Great Pacific Patch’ has brought in immense damage to the marine life, killing thousands of sea birds and turtles along with huge mammals like the blue whale. Even though the industries aim at innovative zero waste missions, it can only reduce up to 70% of the resulting waste [14]. When analyzed about the huge costs to eliminate wastes, it was discovered that the entire problem was centered at improper sorting measures. If plastics were completely diverted from landfills, a potential energy equivalent to 60 million barrels of oils can be saved. The landfills can be reduced to a great extent (About 20%). The sorted recycling yields better results with lesser dependency on raw materials for future developments [14]. There exists innumerous methods to extract and segregate metals like aluminum, copper, and steel along with compostable waste products, like paper and food waste. But the ones which highest degrading factor to the environment, plastic and glass articles, have not yet been segregated with good success rates.

The Use of Modern Technology in Smart Waste Management …

183

Incorporating recognition algorithms with artificial intelligence and machine learning coupled together, can owe to be a great solution for an automated process of post-consumer plastic segregation.

3.4 The System Description The ultimate efficiency is dependent on two main factors. The first being the amount of material segregated by the system. In terms of statistics, this can be well interpreted by Type I error, which is the error of identifying an object incorrectly and not removing it from the trash. The second factor is dedicated to the accuracy of the sorting, which is determined by the Type II error. This indicates the scenario when an object is wrongly detected and placed in another category [15]. The use of Adaptive and Interactive Modelling Systems (AIMS) is to demonstrate the sensor data flow and to output a highly efficient description of object features that completely define the segregation strategies. The system is designed to tackle the challenges of real time data acquisition from the sensors and deliver the information about the machine sensed and learned processing speeds [16].

3.5 Artificial Intelligence Based Sorting Techniques The sorting of waste materials is highly essential to give way to recycling. There are multiple ways to sort the waste materials which significantly determine the cost efficiency and the output levels. The Source segregation involves the separation of recyclable materials into the corresponding different criteria, right at waste production site. It is considered as the best mode of waste segregation by manual self-awareness schemes. The source segregation mode of waste processing show the possibility of waste recovery has the potential to more than 25% with steadily increasing positives. On the other hand, for the fully mixed waste to be processed is labor intensive conventionally. Automating this process is an excellent use of the technology and skill in order to prevent the human interference and the corresponding vulnerability to respiratory and skin prone diseases. These lie on the higher end of financial investments with a waste recovery rate of mere 15% [17]. The machine learning induction algorithm that we propose to use here is capable of changing the conventional methods and efficiency rates, to a better output (Fig. 6). The machine learning induction algorithm is capable of adapting to the changing situations and incorporating new conditions and discoveries with those preloaded findings. One of the major advantages of this lies in the possibility and potential to combine even unrelated and discrete sensor data into the process of decision making. With time, the performance of the inductive system can exceed and go beyond that of an expert and super intelligent system which might be constrained by the hard coded and less acquired knowledge base [15].

184

P. K. Gupta et al.

Fig. 6 Schematic of the waste stream segregation waste [20]

The figure depicts a conceptual understanding of the prototype proposed as the Waste segregation system, its components and the way they interact and communicate with each other. The Sensor interrogator and the database assimilator consists of the PC-Type computers which is handled by a single computer capable of performing the tasks independently. The configuration illustrates the mode of learning by the system. The machine is trained to detect glass, metal articles and plastic, using a variety of possible shapes, deformation levels, sizes, color, and different levels of contamination. Each substance is made to pass through a systematically placed array of unique sensors. Their corresponding responses are recorded and a combined database is established which possesses the accumulated sensor data and the interpreted information. All the containers will now be analyzed and checked using the AIMS. The performance parameters set by the user will determine the comparison and use of different induction algorithms. As soon as the proposed models have been developed and the sensor data will be sent to the interrogator for assessment of the accuracy and stability [18].

The Use of Modern Technology in Smart Waste Management …

185

A machine learning algorithm is designed to learn the ways to perform a given task based on the preloaded data and preset examples. This is referred to as example driven learning. It has the capability of learning on its own through feedback it takes from the environment based on its direct interaction. This is known to be the goal driven aspect of machine learning. We would like to discuss the example driven learning that requires the user to set the list of possible scenarios as examples during the training of the machine code. The examples presented to the machine is usually a vector consisting of input variables along with the output variables. The learning algorithm is designed to produce a function or the model that is capable of mapping the input variables to the output variables [17]. The physical attributes in this waste segregation model is the specifications of the container, such as the size, color, acoustic features and the optical density. This acts as the input variable to the model. Other possible input variables that can be used to describe the container is the details regarding the objects origin and the date of trash collection. These act as a direct link to the system output. The output is the determination of the waste segregation bins to which each of the waste object is to be sent. After feeding the machine with adequate number of examples, the example driven learning algorithm is utilized to create a classification function. The type of the learning algorithm and the related parameters is decided by the system. This is essential for the desired accuracy, speed and the stability of the output [19].

3.6 Overview of the Induction Algorithm and Selection Criteria The AIMS is a highly flexible and powerful system which can operate on both goal and example driven learning. It is known to be the meta-learning algorithm as it has the potential to decide on the right learning algorithm and the corresponding parameter settings with high levels of optimization techniques. This is achieved by a search tool that finds out the one learning algorithm with maximum positivity to the objective of the user. This optimization technique is useful for the fulfillment of one objective. In case of multiple objectives the AIMS chooses a set of models that are Pareto optimal, which is relatively the best to the given objectives. The main advantage of the classification based Pareto optimal models is that these have high levels of flexibility and the perfect idealized behavior under different circumstances. For example: Let us set the objectives to the AIMS as, High accuracy, Minimal time, and minimal use of the sensors.

186

P. K. Gupta et al.

Once the Pareto models are considered, there might be a variation with respect to the efficiency and the precision of the output. For best results, the user could opt for a slower but accurate processing. On a similar line, the volume of the input waste can increase to which a faster and a less accurate model can be employed with faster conveyer movement. If the input sensor undergoes repair or damage, the model is designed to work without the data from that particular sensor.

3.7 Results In order to exhibit the practical working and the corresponding understanding of the working model, we have defined two sorting tasks. The first task involves the determination of the bottles based on the light sensor data. The second task was designed to classify the container into metal, glass and plastics. This is based on the sensor data obtained from a single acoustic sensor [20]. Process Discussion: The Bottle sorting techniques involved the process of testing the bottle quality. Each bottle had to pass through a solar cell and a red Led, whose light had to be obstructed from reaching the cell. The corresponding voltage fluctuation created a waveform of 150 samples per millisecond. This is the input vector and the corresponding output vector was to indicate 0 or 1. 0 corresponds to unclear bottle and 1 to a clear bottle. The error of random guess of the clear substance would be about 40%. The acoustic sensors work on the Fast Fourier Transforms that operate on the input signal. The materials were segregated into metals by recording the sound waves. The FFf calculated and formulated the first 200 harmonics of the power spectrum with a specified bandwidth. These were the input parameters of the object. In order to classify the waste into three different classes, the output vector is composed of Boolean numbers: (o1, o2, o3) in which the metal, glass and plastic corresponded to (1, 0, 0), (0, 1, 0) and (0, 0, 1) respectively. The error of random guess would be about 33% and the error of classifying all clear objects as glass would be 42%.

3.8 Comparison of Existing Learning Algorithms With the explanation on AIMS techniques, we have discovered 3 other learning algorithms which effectively bring down the predictive errors. They are: Simple Recursive Splitting (SRS), Neural Networks (NN), and the instance based (IB) (Table 1).

The Use of Modern Technology in Smart Waste Management …

187

Table 1 To demonstrate the results obtained through various models [20] Experiment type

Learning algorithm

Classification error (%) Mean

Std. deviation

Glass sorting

SRS

20.5

NN

33.8

Material sort

Evaluation time (ms)

Formulation time (s)

1.5

0.41

115

3.4

7.96

253

IB

22.9

1.6

9227

0

SRS

32.7

2.7

0.43

198

NN

7.3

1.6

8.67

88

IB

12.3

1.6

2046

0

4 Conclusion The use of intelligent automation techniques in sorting the waste materials is highly feasible and time efficient. The Control systems have derived the knowledge of machine learning induction algorithms to enhance system adaptability. The preliminary sensor data from acoustic tests and light transmission have the potential to differentiate clear and unclear objects. Apart from this they possess the ability to wisely segregate glass, metal, and plastics using emitted sound waves. Electromagnetic sensors are to be implemented for preliminary metal sorting. Many companies have been establish with waste recycling robots that have revolutionized this field of machine learning. Some of the potential companies with valid investment in technology and objectives of automated waste cleaning are Zen Robotics, BioHitech and Intel.

References 1. Kumar, N.S., Vuayalakshmi, B., Prarthana, R.J., Shankar, A.: IOT based smart garbage alert system using Arduino UNO. In: IEEE Region 10 Annual International Conference, Proceedings/TENCON (2017). https://doi.org/10.1109/TENCON.2016.7848162 2. Gutierrez, J.M., Jensen, M., Henius, M., Riaz, T.: Smart waste collection system based on location intelligence. Procedia Comput. Sci. (2015). https://doi.org/10.1016/j.procs.2015.09. 170 3. Mitton, N., Papavassiliou, S., Puliafito, A., Trivedi, K.S.: Combining cloud and sensors in a smart city environment. Eurasip J. Wirel. Commun. Netw. https://doi.org/10.1186/1687-14992012-247 4. Sinha, T., Kumar, K.M., Saisharan, P.: Smart dustbin. Int. J. Ind. Electron. Electr. Eng. (2015) 5. Vicentini, F., Giusti, A., Rovetta, A., Fan, X., He, Q., Zhu, M., Liu, B.: Sensorized waste collection container for content estimation and collection optimization. Waste Manag. (2009). https://doi.org/10.1016/j.wasman.2008.10.017 6. Kim, B.-I., Kim, S., Sahoo, S.: Waste collection vehicle routing problem with time windows. Comput. Oper. Res. https://doi.org/10.1016/j.cor.2005.02.045 7. Hannan, M.A., Arebey, M., Basri, H., Begum, R.A.: Intelligent solid waste bin monitoring and management system. Aust. J. Basic Appl. Sci. (2010). https://doi.org/10.1007/s10661-0101642-x

188

P. K. Gupta et al.

8. Spira, P.M., Pan, A.: On finding and updating spanning trees and shortest paths. SIAM J. Comput. (1975). https://doi.org/10.1137/0204032 9. Petit, J.: Experiments on the minimum linear arrangement problem. J. Exp. Algorithmics (JEA) (2003). https://doi.org/10.1145/996546.996554 10. The traveling-salesman problem: Mathematics in Science and Engineering (1977). https://doi. org/10.1016/S0076-5392(08)61182-0 11. López, J.G., Imine, M., Rumín, R.C., Pedersen, J.M., Madsen, O.B.: Multilevel network characterization using regular topologies. Comput. Netw. https://doi.org/10.1016/j.comnet.2008. 04.014 12. Hunt, V.D., Puglia, A., Puglia, M.: RFID-A Guide to Radio Frequency Identification. https:// doi.org/10.1002/9780470112250 13. Ali, M.L., Alam, M., Rahaman, M.A.N.R.: RFID based e-monitoring system for municipal solid waste management. In: 2012 7th International Conference on Electrical and Computer Engineering, ICECE 2012 (2012). https://doi.org/10.1109/ICECE.2012.6471590 14. Torres-García, A., Rodea-Aragón, O., Longoria-Gandara, O., Sánchez-García, F., GonzálezJiménez, L.E.: Intelligent waste separator. Computacion y Sistemas (2015). https://doi.org/10. 13053/CyS-19-3-2254 15. Tcheng, D.K., Lewis, M.R., Newell, T.A.: Applications of machine learning to sorting of recyclable containers. In: 215 Conference on Solid Waste Research in the Midwest, Chicago, Illinois, Oct 1990 16. Lewis, M.R., Newell, T.A.: Development of an automated clear/color sorting system for recycling containers. In: Conference on Solid Waste Research in the Midwest, Chicago, Illinois, Oct 1990 17. Tcheng, D.K., Lambert, B.L., Lu, S.C.-Y., Rendell, L.A.: Building robust learning systems by combining induction and optimization. In: Proceedings of the International Joint Conference on Artificial Intelligence, San Mateo, California. Morgan Kaufmann (1989) 18. Apotheker, S.: Garbage In, But What Comes Out? Resource Recycling, October 1990 19. Tcheng, D.K., Lambert, B.L., Lu, S.C.-Y., Rendell, L.A.: AIMS: an adaptive interactive modeling system for supporting engineering decision making. In: Proceedings of the Eighth International Workshop on Machine Learning, San Mateo, California. Morgan Kaufmann (1991) 20. Tarbell, K.A., Tcheng, D.K., Lewis, M.R., Newell, T.Y.A.: Applying machine learning to the sorting of recyclable containers. University of Illinois at Urbana-Champaign, pp. 209–215, Urbana, Illinois

Video Streaming Communication over VANET Gurpreet Singh

Abstract Recent evolution in wireless technologies and the Internet of things is promoting the design of new communication systems such as in Vehicular Ad Hoc Networks (VANETs). These networks are systems to offer new infotainment approaches and traffic safety services with smart devices in smart surroundings. A Smart City is intended as an environment which is managed and monitored with latest technologies and new types of communication in order to hike the quality of life with ingenious services. Vehicles are already provided with a variety of computational devices that manage or assist drivers in many jobs such as localization, safely crack, parking, passenger’s pleasure and etc. [1]. One of the major problems facing smart cities communication is to ensure well organized services delivery under economic, community and environmental situations. Hence, multimedia communication and notably video streaming are forecast to be of high convenience for traffic management as well as for providing enjoyment and advertising facilities. With video streaming services, real-time information will be engaged and supplied by vehicular networks for safety and effective goals. In this chapter we focus on video streaming transmission over vehicular ad hoc networks for facilitating passengers.

1 Introduction to VANET Intelligent Transport System (ITS) is the advanced transport system. ITS is first introduced to increase the road protection by engaging Vehicle-To-Infrastructure (V2I) and Vehicle-To-Vehicle (V2V) communication on the road. New information and communication technologies are registered in the field of road transport, traffic configuring and mobility management, etc. Besides the scope of road safety, the licenses-free VANET communication is also encouraging way to deliver many services to upgrade the driving experience for users on vehicles, such as traffic observation, driving help and infotainment. By exploiting Road Side Units (RSUs) along G. Singh (B) Department of Computer Science, Baba Farid College, Bathinda, Punjab, India e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_12

189

190

G. Singh

Fig. 1 VANET architecture

the road, such services are competent for the vehicular drivers even in the most rural area. In VANET, vehicles are considered as nodes. There are two kinds of communication are possible in VANET that is vehicle to vehicle (V2V) and vehicle to infrastructure (V2I) [2, 3]. The communication facilitates the vehicles to transfer messages within this infrastructure as shown in Fig. 1. The infrastructure includes all protocols and hardware equipments for transmission of information. VANET is providing number of features to facilitate users and for their road safety. There are some unique characteristics of VANET [4]. High Dynamic Topology: The speed and choice of path defines the dynamic topology of VANET. According to assumptions two vehicles moving away from each other with a speed of 60 km/h (25 m/s) and if the transmission range is about 250 m, then the link between these two vehicles will last for only 5 s (250 m). This defines its highly dynamic topology. Frequent Disconnected Network: The above feature necessitates that in about every 5 s or so, the nodes needed another link with nearby vehicle to maintain seamless connectivity. But in case of such failure, particularly in case of low vehicle density zone, frequent disruption of network connectivity will occur. Such problems are at times addressed by road-side deployment of relay nodes. Mobility Modeling and Prediction: The above features for connectivity therefore needed the knowledge of node positions and their movements which is as such very difficult to predict keeping in view the nature and pattern of movement of each vehicle. Nonetheless, a mobility model and

Video Streaming Communication over VANET

191

node prediction based on study of predefined roadways model and vehicle speed is of paramount importance for effective network design. Communication Environment: The mobility model highly varies from highways to that of city environment. The node prediction design and routing algorithm also therefore need to adapt for these changes. Highway mobility model is essentially a one-dimensional model which is rather simple and easy to predict. But for city mobility model, street structure, variable node density, presence of buildings and trees behave as obstacles to even small distance communication make the model application very complex and difficult. Unlimited Transmission Power: The node (vehicle) itself can provide continuous power to computing and communication devices. Hard Delay Constraints: The safety aspect (such as accidents, brake event) of VANET application warrants on time delivery of message to relevant nodes. It simply cannot compromise with any hard data delay in this regard. Therefore, high data rates are not as important as an issue for VANET as overcoming the issues of hard delay constraints. Interaction with onboard Sensors: These sensors help in providing node location and their movement nature that are used for effective communication link and routing purposes. Higher Computational Capability: Indeed, operating vehicles can afford significant computing, communication and sensing capability. Rapidly Changing Network Topology: Due to high node mobility, the network topology in VANET tends to change frequently. Potentially Unbounded Network Size: VANETs could involve the vehicles in one city, several cities or even a country. Thus, it is necessary to make any protocols for VANET is scalable in order to be practical. Anonymous Addressee: Most applications in VANETs require identification of the vehicles in a certain region, instead of the specific vehicles. This may help protect node privacy in VANETs. Time-Sensitive Data Exchange: Most safety related applications require data packet transmission in a timely manner. Thus, any security schemes cannot harm the network performance of VANETs. Potential Support from Infrastructure: Unlike common MANETs, VANETs can actually take advantage of infrastructure in the future. This property has to be considered to make any protocols and a scheme for VANET is better.

192

G. Singh

Abundant Resources: VANET nodes have abundant energy and computation resources. This allows schemes involving usage of resource demanding techniques such as ECDSA, RSA etc. Better Physical Protection: VANET nodes are better protected than MANETs. Thus, Nodes are more difficult to compromise which is also good news for security provisioning in VANETs. Partitioned Network: Vehicular networks will be frequently partitioned. The dynamic nature of traffic may result in large inter vehicle gaps in sparsely populated scenarios and hence in several isolated clusters of nodes.

2 Introduction to Video Streaming in VANET Video streaming is one type of data flow. Video streaming support over VANETs is used to increase the efficacy of exigency action in case of any mishap, streaming a live video from the mishap place and broadcasting the information to other vehicles. Video streaming poses stringent requirements specially regarding successful delivery ratios and end-to-end delay. Video streaming consists of streaming video content in a close to real-time manner in which delay cannot exceed a few seconds. It is also fundamental to provide solutions that deliver the vast majority of the content transmitted since error correction mechanisms that requires interaction between receivers and transmitters may deteriorate prohibitively the experience end-to-end delay. VANETs are particularly a challenging scenario for video streaming due to its highly dynamic topology, shared wireless communication medium and bandwidth constrains. The combination of VANETs’ challenges to the requirements of video streaming requires the investigation of novel solutions for packet delivery [5]. A architecture view of video streaming in vehicular IoT domain is shown in Fig. 2 focusing on encoding-decoding layers. The benefit of systematic streaming is that, the user does not need to waste time in downloading a video content for playing the video. Real-time communication across wireless automobile networks facilitates more wide-ranging information via multimedia such as video which can help drivers to make more acceptable decision within catastrophe area. The allocation of video streaming to driving vehicles can achieve very pretty applications. For instance, in highway protection system, the transmitted urgent situation message allows Replier better understanding of the nature of the problem about the catastrophe area. In this type of application the video source can be a vehicle or a base station (access point) with suitable capability while the video recipients are vehicles that are travelling in the domain area. There are two approaches for video streaming in VANET:

Video Streaming Communication over VANET

193

Fig. 2 Architecture view of video streaming in vehicular traffic domain

2.1 Infrastructure Based Approach This approach relies on deployed access points. In the context of vehicular networks, infrastructure is a set of specialized communication devices supporting the network operation. Common properties include (but are not restricted to) network centrality, communication bandwidth, storage space, and high availability [6]. Because vehicular network devices are initially envisioned to be located at roadsides, they are commonly referred to as RSUs and may provide a large number of functions, such as the following: (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xii) (xiii) (xiv) (xv) (xvi)

Broadcast Channel allocation Caching Content download Data dissemination Data aggregation Data scheduling Gaming and streaming Gateway Hand-off Vehicles localization QoS Real-time support Routing Security Multihop communication

194

G. Singh

2.2 Infrastructure-Free Approach This approach forwards video file to vehicles in the vicinity zone. It is a decentralized vehicular density estimation technique that dissects each road into a number of fixed size cells based on transmission range of vehicles, as shown in Fig. 2. A vehicle closest to cell center is elected as a group leader that is responsible for estimation of cell density and forwarding the cell density information in form of cell density packet (CDP) to other cells. The group leaders add their cell density to the received CDP and forwards the CDP along the path until it reaches the junction. In this way, vehicular density of each cell is estimated and combined together to find the total vehicular density on a specific road segment. The calculated CDP information is available on the road junctions that plays important role in VANET applications and routing protocols [7]. The first approach costs too much due to high base station deployment and maintenance services. Furthermore, in vehicle to vehicle communications approach it totally relies on the infrastructure-less vehicular network and it supports video applications over vehicle to vehicle communication. This is a more challenging task due to the lack of a pre-installed infrastructure to provide centralized control for the entire networks. On the other hand, the advancement of wireless technology and video compression techniques gives a practical solution for video streaming through VANET. Vehicles have embedded with large storage power and provide high computation capability. The connectivity process between vehicles demand robust techniques installed within each vehicle so that it can be able to connect especially in emergency situation on the road.

2.3 Classes of Video Streaming Over VANET We have divided the study of video streaming over VANETs in two different classes based on figure of receivers. We have studied individually design of solutions for unicast and broadcast, Video Unicast can be described as the transmission of video content from a single source to a single destination while Video Broadcast (or Video Dissemination) as from a single source to all nodes within a distance to it. We have not cross-examined the model of Video Multicast (from one source to many destinations and not necessarily all nodes) but our deliberation regarding Video Dissemination can be used to the growth of a multicast solution that uses video distribution and filters the content to be sent to application layers of only the suitable receivers.

2.3.1

Unicast

Video Unicast over VANETs occurs through the VIdeo Reactive Tracking-based UnicaSt (VIRTUS) protocol. This provision is also extensively assessed and it’s perform is compared to other outputs. It is also a receiver-based result that has the

Video Streaming Communication over VANET

195

choice of relay nodes balanced between geographic advancement and link stability. VIRTUS is a meaningful step towards high rates of successfully and timely transport of video content from one source to one end of line over VANETs [8].

2.3.2

Broadcast

The following are the requirements of VANET protocols when designing a protocol for broadcasting [9]: Scalability: Protocol should well with all terrain such as urban, rural, metro cities, dense and sparse areas. So designing a protocol that suits for all the areas is a challenging task. Effectiveness: The broadcast protocol has to ensure that all the vehicles in the destination region receive the broadcasted message. Efficiency: While broadcasting the messages, redundancy is s bigger issue. The protocol should work as efficient as possible to avoid redundancy of messages. Dissemination Delay: Emergency messages have to be notified immediately without delay. Separate treatment of emergency messages in the network is under research. Delay Tolerant Dissemination: The protocol should be capable of storing the messages for some amount of time when network is disconnected. The protocol has to forward it when new vehicles as connected to the network. Robustness: Broadcast protocol has to deal with packet loss with the purpose of operating accurately in vital safety applications. Broadcasting of real-time video, mostly accidents relevant, plays a major role in Vehicle Ad hoc Networks (VANETs). Multi-view video enables the discernment of the targeted scene from multiple sides. By taking the real-time multi-view video broadcast, drivers can obtain better knowledge of the highway traffic and road states [10]. Broadcasting protocols for VANET can be classified into six broad categories: 1. 2. 3. 4. 5. 6.

Table driven broadcasting Cluster based broadcasting Topology based broadcasting Location based broadcasting Distance based broadcasting Probability based broadcasting

3 Protocols for Video Streaming in VANET 3.1 ADV Routing Protocol The Adaptive Distance Vector (ADV) is a Distance Vector Routing algorithm that uses consecutive numbers to deter loops in the network. ADV is same as other distance

196

G. Singh

vector algorithms but it minimizes the routing load by fluctuating the frequency and size of routing updates owing to flexibility in traffic and node mobility. It keeps routes to active receivers only so that the number or entries publicized is reduced. It activates partial and full updates so that periodic full updates are eliminated.

3.2 AODV Routing Protocol AODV is a distance vector routing protocol. Unlike proactive routing protocols like DSDV, AODV is reactive which means that it only wish a route when required and does not need nodes to keep routes to destinations that are not presently active. When a node wants a route to a destination which is currently unknown or the previous node has died, it broadcasts a RREQ. The source node uses increase rapidly ring search technique to avoid unneeded flooding of RREQs. This technique uses locality to find the destinations. The TTL (Time-to-Live) field in the IP header of the RREQ packet is used to control this search [11].

3.3 Enhanced User Datagram Protocol (EUDP) This is new protocol named Enhanced User Datagram Protocol (EUDP) for video streaming in VANET. Unlike User Datagram Protocol (UDP) which did not examine any recovery mechanism of erroneous packets, EUDP uses Sub-Packet Forward Error Correction (SPFEC), and choose the unequal protection of video frame type.

3.4 Real-Time Transport Protocol (RTP) Real-Time Transport Protocol (RTP) facilitate node-to-node network transport functions suitable for transmission of applications which are delay-intolerant, such as audio, video, or simulation data on unicast or multicast network services. RTP does not provide resource booking and does not assure quality-Of service for real-time services. The data transport is monitored by a control protocol (RTCP) to watch reliable delivery of data in large multicast networks, and to deliver marginal control and give functionality. RTP and RTCP are so implemented that they are independent of the underlying transport and network layers. The protocol assists the use of RTP-level translators and mixers. SDP is visualize to define multimedia sessions for session announcement, session invitation etc. It provides a layout for defining session information to session members. This information constitutes the name of the session and media type and format.

Video Streaming Communication over VANET

197

4 Conclusion In this chapter, we tried to explore VANET communication especially for video streaming. With the introduction to VANET, video streaming techniques has been discussed. The two different classes e.g. Unicast and Broadcast explore the way by which information can be passed within infrastructure from different nodes. At the end, we tried to introduce some types of protocols required for video streaming communication in VANET. Video streaming capacities over VANETs are preeminent to the development of interesting and valuable services. However, VANETs’ highly active topology poses as a market challenge to the fulfillment of video screaming’s conscientious needs.

References 1. Weiland, R.J., Purser, L.B.: Intelligent transportation systems. Trans. New Millenn. (2000) 2. Toor, Y., Muhlethaler, P., Laouiti, A.: Vehicle ad hoc networks: applications and related technical issues. IEEE Commun. Surv. Tutor. 10(3), 74–88, (2008) 3. Hartenstein, H., Laberteaux, K.: A tutorial survey on vehicular ad hoc networks. IEEE Commun. Mag. 46(6), 164–171 (2008) 4. Kaur, M., Kaur, S., Singh, G.: Vehicular ad hoc networks. J. Glob. Res. Comput. Sci. 3(3), 61–64 (2012) 5. Rezende, C.: Video streaming in vehicular ad hoc networks: challenges, protocols and the use of redundancy. Ph.D. Thesis, University of Ottawa (2014) 6. Silva, M., Masini, M., Ferrari, G., Thibault, I.: A survey on infrastructure-based vehicular networks. Mobile Inf. Syst. 28 pp. (2017), Article ID: 6123868 7. Bilal, S.M., Khan, A.R., Khan, S.U., Madani, S.A., Nazir, B., Othman, M.: Road Oriented Traffic Information System for Vehicular Ad hoc Networks. Springer Science Business Media, New York (2014) 8. Rezende, C., Boukerche, A., Ramos, H., Loureiro, A.: A reactive and scalable unicast solution for video streaming over VANETS. IEEE Trans. PP(99), 1–1 (2014) 9. Ramalingam, M., Thangarajan, R.: A study of broadcasting protocols and its performance in VANETs. Int. J. Emerg. Eng. Res. Technol. 4(3), 1–10 (2016) 10. Liu, Z., Dong, M., Zhang, B., Ji, Y., Tanaka, Y.: RMV: real-time multi-view video streaming in highway vehicle ad-hoc networks (VANETs), pp. 1–6 (2016) 11. Masood, O., Akram, A., Majeed, M.: Performance evaluation of ADV with AODV for real-time and multimedia applications in vehicular ad-hoc networks (VANETs). Int. J. Comput. Netw. Technol. 1(2), 119–127 (2013)

An Update on Effective Patient Monitoring and Diagnosis of Diseases in the Current Scenario Sindhu Rajendran, Meghamadhuri Vakil, Praveen Kumar Gupta, Lingayya Hiremath, S. Narendra Kumar, Ajeet Kumar Srivastava and Vidhya Shree

Abstract Patient monitoring is always a challenging task. In this sense, a vast country like India, it becomes an uphill task in order to identify the disease in a better and faster way in each and every part of the country. In the modern scenario, the development of some useful wearable electronics and their uses in IOT based technologies have shown a promising future in terms of disease diagnosis and associated monitoring. The uses of wearable devices and mobile apps now have been integrated with telemedicine and telehealth efficiently with respect to monitoring diseases mainly associated with cardiovascular system, central nervous system, blood pressure, diabetes and lung associated disorders. The uses of mobile apps in android based cost effective devices have given a revolution in terms of faster and cheaper disease diagnosis and related researches. In recent times, the concept of M-Health (Mobile health) and E-Health (health care supported by ICT) have been useful in improving and assisting people’s health, one such Device is AliveCor heart monitor which is a mobile phone based electrocardiogram. Remote wearable ECG monitoring and Mbraintrain smart devices have received a great attention in the medical era. This chapter enlightens the basic methodology and uses of various wearable health monitoring devices which may or may not be associated with IOT. Keywords Internet of things (IOT) · Wearable devices · M-health · E-health

S. Rajendran (B) · M. Vakil Department of Electronics and Communication, R. V. College of Engineering, Bangalore 560059, India e-mail: [email protected] P. K. Gupta · L. Hiremath · S. Narendra Kumar · A. K. Srivastava Department of Biotechnology, R. V. College of Engineering, Bangalore 560059, India V. Shree Department of Electronics and Instrumentation, R. V. College of Engineering, Bangalore 560059, India © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_13

199

200

S. Rajendran et al.

1 Introduction Wearable technology is a booming technology in patient’s health monitoring. It is a small, smart electronic device which has sensor and it is wearable on body as well as on clothe. In healthcare monitoring Wearable devices have been implemented, which measures critical parts of the body. Most widely used among the majority of devices are electrocardiogram (ECG), temperature measurement and electroencephalogram (EEG). There have been recent advancements in wearable devices for vital signs measurement. Wearable devices such as AliveCor heart monitor, Mbraintrain smart devices and Remote wearable ECG monitoring helps in exchanging data through internet with the concerned division without human involvement. Wearable technology is growing technology in many fields and has many applications. There are many commercial uses of this technology such as increasing demand of smart watches and fitness bands to track ones activity. Wearable technology is not only growing in commercial side it also has applications in monitoring health, in textiles and in navigation systems. The age of wearable electronics has made our lifestyles simple. It is in fact an implementation of Internet of Things (IoTs) in a smaller domain integrating your watch or wrist band to the analytics of your smart phone. It is just a simple interconnecting platform for your logs, records and control data. With the corresponding mobile application, these wearable sensors impart the technical sensed input that is put across to us in rational outputs. Since they have the access to and from the internet, the smart wearable’s form the integral part of health and fitness monitoring. One of the best integrated and highly successful smart wearable is Google Glass. With an optical head mounted display, Google glass is just a super intelligent computer in the form of a pair of glasses. It works like a phone with virtual reality. The smart watches nowadays have inbuilt body sensors that is integrated with GPS. Now you can track the distance you walk, the heartbeat, the corresponding calories burnt and thereby keep a check on your health. Many sensors are based on blood glucose meters which monitor your diet and precisely calculate the about the insulin dose required as per the diet intake. This is very essential for diabetic patients who need a regular dose of insulin in form of injections. The smart wearables are also designed with inbuilt fitness and health trackers to remind you to drink water and take a break from work to reduce stress levels. E-textiles are another domain gaining momentum over the years. Apart from being excellent clothes which can be washed and used as normal ones, they possess smart sensing capabilities. They monitor your body temperature with respect to the external humidity and pollution levels to suggest you the best routine to be followed. With the introduction of IoT in healthcare, along with the traditional tools used in hospitals for monitoring, diagnosis and therapy will be replaced with new intelligent and programmed systems, in-home health monitoring can also be implemented in a more broad way. In the end of 20th sanctuary there were proofs that patients in hospitals are dying because of less attention. In order to overcome this problem and to find an intelligent system to recognize the deterioration in patients, the first early

An Update on Effective Patient Monitoring …

201

warning score (EWS) was introduced in 1997. The EWSs is used as a guide to reduce the risk of illness and deterioration in patient [1]. Hence reducing redundant harm. This guide uses parameters such as temperature, heart rate/pulse rate, blood pressure, oxygen dissemination, respiratory rate and AVPU response to observe the patient’s health throughout their stay in hospital and recognition of fading, characterized as death or admission to intensive care unit (ICU). EWSs are computer systems which develops an algorithm in order to keep the record of critical signs of patient’s health by hand or automatically. There are several EWSs consistently used in countries such as USA, UK, Netherlands and Australia. After the EWSs algorithm, the concept of self awareness is necessary to inculcate systems with the required tools to obtain the dynamic changes with respect to the characteristic and the corresponding interest. These are mainly the readability, optimality and adjustability. Many of these concepts are of significant interest for almost a sum of 26 billion estimated devices that are known to be connected by the broad domain of Internet of things (IoT) favourably by 2020. Therefore, it is quite essential to look for the various applications using the concept of self-awareness. The area of research is extended to mobile applications, networks, patient health monitoring systems and cloud computing [2]. This is the ultimate motivation behind the innumerous benefits which can be obtained using smart wearables. One of the dominating and pioneering architectures of self-awareness is the Observe-Decide-Act (ODA) loop is shown in Fig. 1. Observe is used for internal and external data which is first retrieved from the collected data through sensors [3]. This network is pre-processed and is therefore termed observe. The next situation awareness coupled with self-awareness increases to assess and process the observations. The best configuration of the system is decided and implemented for the system benefit, and is referred to as decide. This can be speculated in two different scenarios: the foremost part helps you evaluate the health of the subject appropriately, irrespective of the noise or the incorrect values. The important part of this configuration is the improvement for the system operations. The main parameter which can be controlled is the ‘Attention’. This is completely based on the observations followed by the decisions adopted. The Attention, determines certain parameters that is related to the complete activity of the sensors. The output of these sensors is converted to the sensor comprehensible signals in the configuration unit. Once, it passes the sensor network, it is essential to note that the critical part of the entire framework is the model and its environment. A sensor designer can implement his or her sound technical knowledge for a comprehensible model with a high reality quotient. The complementary sensor collected data is not sufficient to produce a full image unless the complete user feedback is given to the system. This feedback given to the system is essential for the remote device and the support system. Each of these methods is essential to build a sustainable and worthy product for the environment.

202 Fig. 1 Architectures of self-awareness

S. Rajendran et al.

Observe

Decide

Act

2 Current Scenario of Patient Health Monitoring System In most of the countries there is an increase in mortality rate due to chronic diseases thereby leading to high health care costs. Heart disease has been a major widespread public health problem and costly for healthcare systems. Another common chronic disease is Diabetes which is one of the most common metabolic diseases. Diabetes care may account for up to 15% of national health care budgets. The quality of care can be improved drastically by providing high quality monitoring devices with nutrition program and physicians coordination. Modern technology tools have been introduced to reduce the cost of health sectors. There is a need for continuous assessments and compliance with self management programs for prevention of chronic diseases. One of the major tools used is Telemedicine for remote health care which can be easily accessible to health information thereby reducing the time and travelling distance [4]. M-Health systems are used for medical diagnosis and monitoring system which provides doctors access to the database to get patient’s information (Medical history if any) so that the treatment is easier especially in emergency situations so that there is no delay and error in suitable treatment.

2.1 Concept of M-Health and E-Health M-Health also called as mobile health is used for public health supported by mobile phones (smart phones) and for other application in medical care system. The application of M-Health is to exploit the mobile phones and communication strategy to instruct clients about preventive health care services as well as in collecting health data and real time monitoring of patient critical signs. It also has other applications such as surveillance of diseases, epidemic outbreak tracking and chronic disease management. M-Health field has emerged as a sub-section of E-Health, which uses information and communication technology such as mobile phones, computers, patient monitors, etc., for health services and information. Table 1 shows the E-health device and its purpose. E-health is a healthcare practice supported by electronic processes and communication. It specifies the narrow definition covering the electronic/digital processes in health using internet. Development of E-health solutions that can withstand the growing threats in cyber-security.

An Update on Effective Patient Monitoring …

203

Table 1 E-health device and its purpose E-health devices

Purpose

Glucose/BP/HR monitor Personal health record

Improve personal monitoring Engage patient in self care Data for health professional

Pedometer GPS tracking

Monitor exercise levels Encourage wellness Population: level of activities

Laboratory tests Electronic health record

Diagnosis/manage individuals Track disease manifestations Resource allocation

3 Methodology for Patient Monitoring The Table 2 specifies the different detectable indicators in health monitoring system. A series of physiological parameters can be used for measuring the overall health of the patient. This could be the body motions with respect to the hands, limbs and foot, which can be used for determining the training and fitness levels. The skin temperature is an indicator of various other factors of general health and the stress and anxiety levels. The Heart rate is the most important parameter which sets the basis for the determination of the health status.

Table 2 Different detectable indicators in health monitoring system [8]

Indicator

Position

Measured parameter

Possible application

Motion

Hand and limb

Strain and pressure

Rehabilitation and gesture identification

Face and throat

Strain

Expression and phonation detection

Respiration

Chest and nostrils

Strain, pressure, humidity

Detection of cardiac arrest, apnea, emotional control

Heart rate

Chest, wrist, neck and fingertips

Strain

Detection of heart failure and cardiovascular disease

204

S. Rajendran et al.

1. Motion sensors: In Human-Machine Interface (HMI) devices, the hand gestures make the entire device a success. With finger sensors now embedded in smart phones, the smart technology it taking a lead in our essential part of life. Flexible strips which have elastic properties are in use to monitor the fingers and the associated motion of the hand [5]. The principle of working is that the motion induces a finger-curvature variation that is manipulated to the corresponding electrical signals. The obtained electrical parameters help determine the diagnosis of Parkinson’s disease. This is also used in popular Sign Language devices to decrypt visual sign gestures. In order to extend the motion detection to hands, limbs, elbow, knee, shoulders, etc., we would need larger stretchable bands with the corresponding upgrade to the detecting device [6]. To increase wear ability, these fingers and hand gesture instruments come with glove like sensor devices. The sensor strips are attached to the finger positions, in order to detect the various motions associated with the hands: making a fist, hand gestures, direction of the pointing finger, etc. This finds immense use in robotic control of arms and translation of Morse code. This is now gaining huge success rate with high accuracy. The human motion detecting devices are very essential in secure places like the borders of nations, in nuclear power plants, hydropower dams, etc. Another significance of these motion detectors are in the gaming field, to bring in augmented and virtual reality highly realistic, through control by real action of the player. 2. Face and Throat: The facial expressions are in fact the only drawback in the robots or the highly intelligent devices. The human face express nearly thousand different feelings which is not clearly conveyed through words. In order to recognize facial expressions is a must to develop highly accurate Human Machine Interface (HMI) devices. In order to bring in the reality inside a virtual assistant, it needs to read and interpret the various facial expressions. The principle used in this is the use of facial muscles which are responsible to depict different facial gestures. The broad classification of the 9 facial expressions is essential to sort in the input muscles contractions and relaxations into the corresponding message to be directed [7]. The tape like sensors are attached to the major and significant parts of the face like the forehead, cheeks, chin and the sides of the eyes, (Canthus, Philtrum, Angulus oris), which detects the changes in the corresponding muscles. Voice can be easily lost, in cases of accidents and this result in the loss of clear speech. Voice recognition and artificial voice are the most innovative solutions for them. Even a person born dumb, can be given the boon of voice based on the muscle movement in the throat. Since different vocalizations indicate different muscles which are either strained or relaxed, we need to work on the calibration of the sensors to note the differences and tally it to a particular voice notation. The strained sensors in different muscle

An Update on Effective Patient Monitoring …

205

changes will alter the electrical parameters, like resistance or capacitance to convert it into electrical signals. In order to capture even the smallest vibrations, the strain sensors need to be extremely sensitive. With latest developments in this field, the integration of many languages has been possible. Thus, the detected muscle changes and vibrations in the throat can be implemented as an artificial intelligent throat with an electronic voice. 3. Breath: Irregular breath patterns are an indication of various respiratory problems. The breath sample from the inhaled and exhaled breath condensate can be assessed for the changes and problems like, sleep apnea, anemia, asthma, and pulmonary diseases. Figure 2 shows the detection of breath and SpO2. The breath sample can be obtained with the sensor integrated face mask, which has inbuilt temperature sensors with the relative volumes. The breath efficiency is determined with SpO2, whose optimized analysis is obtained by the flexible optoelectronic device that can be mounted on finger tips [8]. 4. Heart rate:

Fig. 2 Detection of breath and SpO2 [8]. a Detecting breath with a sensor embedded in mask, and the curve shows the results under normally breathing (1), taking a deep breath (2), paused (3) and randomly breathing (4). b Detecting breath with a strip sensor worn in the chest. c Detecting the SpO2 with a sensor attached on fingertip

206

S. Rajendran et al.

Heart rate can indicate almost all problems faced in the body, with underlying causes. The heart beat rate and the pulse is simply detected by the strain the corresponding pressure sensors in wrist, chest, sides of neck, and finger tips. The ECG is responsible to detect the possibility of heart attacks with a definite ratio [9]. The electrodes which play a vital role in detecting these signals for the ECG, can be obtained from the thorax, ankle and essentially the wrist. In recent studies, it has been proved that pencil based electrodes can detect the ECG signals better than the usual Ag/AgCl ones. The bipolar ECG sensors, along with other medical biosensors can establish a complete patient health monitoring system with utmost efficiency. Heart Rate Control: System and Overview [10]. Figure 3 shows the block diagram of the heart monitoring system. The system is designed to hold the patients temperature and heart rate recording for long period of time. The Health monitoring system uses the sensors which can collect health related data for data acquisition [11]. The communication is done by the controller for collecting the data on the internet. To get health related information in a comprehensible format, it is displayed on the web page which is the data management. The following components are used in the heart rate monitoring system: 1. Arduino Uno: the microcontroller used is the Arduino Uno which is based on the Arduino IDE software. 2. The Temperature sensor used in the working model is LM35 which measures the body temperature and the sensors when I contact with the body senses its temperature. It is accurately calibrated linearly to measure in terms of Celsius. It has a low self heating capacity that makes it independent on its own, without the need of frequent calibration. 3. Pulse Sensor: when a finger is placed on the sensor, it is designed to give the analog output of the heart rate. As it works, the LED placed on the top side will blink with each heart beat. To see the sensor output, output pin of the sensor is under the control of the controller. The Light modulation by blood flow determines the working of the nerves at each heart pulse.

ESP 82666

Things peak (IOT PLATFORM)

Arduino Uno

LCD

Temperature Sensor Pulse Sensor

Fig. 3 Block diagram of heart monitoring system

An Update on Effective Patient Monitoring … Fig. 4 Flow diagram of the health monitoring system

207

Sensing

Processing Unit

Cloud Server

Analysis

4. WiFi Module: The ESP8266 wifi module is a self contained SOC derived from the IP/TCP Protocol stack which offers the controller the access to wifi network. 5. IoT Platform: You can use Thinkspeak to send any kind of data from any internet supporting device to the cloud. It can be configured to relate to actions and alerts based on the real time data. This gives a wide platform for those who wish to enable it to easily capture sensors data and convert it into appropriately useful information. Flow Diagram of the System: Figure 4 shows the flow diagram of the health monitoring system. The Results obtained from the sensor data collection is analyzed and if there is any abnormal behavior, an emergency plan is worked out to immediately inform the doctor about the patient’s health. This reduces the Critical conditions faced in the hospital.

3.1 Sensors Used in Patient Health Monitoring System 3.1.1

E-Health Sensor Platform

As an alternative to the costly and high worth medical business, E-health sensor is used to monitor the health parameters of the patients. It gives an economical and open source environment in the medical care system. The medical analysis like ECG, EMG, body temperature, galvanic skin and pulse, can be used to detect the variations over time. These sensors are linked to the patient monitoring systems which compare readings to accordingly use the shield. Embedded boards like Raspberry Pi, Arduino and BeagleBone Black are used to store and process the data shield. This can be accessed by the computer or the external power supply. If the module faces any issues during implementation, we can make use of the outer force supply (12 V–2 A) on the board, when the USB ports are not sufficient [12] (Fig. 5).

208

S. Rajendran et al.

Fig. 5 E-health sensor shield [13, 14]

3.1.2

Arduino Uno

The Arduino Uno is the ATMega328 microcontroller which incorporates 14 digital pins as the input. It uses 6 analog productions as the output, along with the ICSP header, USB port and reset button. This board is completely equipped to be interfaced by the PC using a USB link. The force source to power the Arduino can be from a USB or from the external power supply. The connector is usually a 2.1 mm center-positive plug that rightly fits into the board’s energy jack. To program the device board an integrated Development Environment (IDE) is used. The Arduino IDE 1.0.1 collects the data from the sensor and shows the detected worth on PC [13, 14] (Fig. 6).

Fig. 6 Aurdino Uno version R3 board [13, 14]

An Update on Effective Patient Monitoring …

209

Fig. 7 Glucometer sensor [13, 14]

3.1.3

Glucometer Sensor

To determine the estimated blood sugar ratio, the glucometer is used as clinical equipment. A little drop of blood is placed on the biostrip which uses intensive sensor technology to detect the glucose levels accurately. The glucose level is indicated at the display in mg/dL [15] (Fig. 7).

3.1.4

Airflow Sensor

The respiratory variations over the respiratory rate are essential for the early cautionary detection of apnea and hypoxemia [16]. The Airflow sensor is needed to gauge the rate of breathing and quality of the exhaled breath. It is essential for the determination of troubles in breathing of the nose or even mouth (Fig. 8).

3.1.5

Patient Position Sensor

it is essentially an accelerometer which observes the patient’s situation with respect to standing, sitting, prone, supine, left and right supine). With regular supervision, one can easily detect the sleep apnea diseases followed by the troubled leg syndrome [17]. The quality of the sleep analysis is totally dependent on the movement analysis

210

S. Rajendran et al.

Fig. 8 Airflow sensor [13, 14]

Fig. 9 Patient position sensor [13, 14]

of the sleep patterns. This particular sensor finds immense use for elderly patients to monitor their fall or syncope (Fig. 9).

3.1.6

ZigBee:

a customization of low energy consumption coupled with low prices, it is built with the mesh networking technology standard. The remote control with observing uses in diverse fields can be used to give high dependability and greater coverage of the

An Update on Effective Patient Monitoring …

211

specified zone. The range of operation of the ZigBee is from 10 to 100 m radius. This innovative idea is estimated to be less complex and comparatively cost efficient with respect to Bluetooth. A long battery life, with a low information rate, and secure systems has a characterized rate of 250 kbps with a fluctuation over the transmission rates from 20 to 900. In order to exchange information from the transmitter station to the receiving station, XCTU programming is used to program the ZigBee. Before it begins to communicate between the transmission and receiving stations, it first needs to be programmed accordingly [18].

4 Wearable Devices Some of the most widely used wearable devices are listed below: 1. Wireless blood pressure monitor: Nokia-BPM+ is a wireless blood pressure monitoring which is elegant, minimal design compact health device which synchronizes with your mobile automatically and using the apps the measurements can be tracked (Fig. 10). 2. Groundbreaking Smartphone Ultrasound Device: MobiUS SP1 system is used by professionals to diagnose people who are in rural places especially for monitoring pregnancies, abdominal, cardiac, peripheral vessel scans etc.

Fig. 10 Wireless blood pressure monitor (Google pic)

212

S. Rajendran et al.

3. Remote Cardiac Monitoring System: This healthcare device is used to monitor cardiac arrhythmias by sending data to the service’s cloud platform, it is capable of tracking heart rate, activity level and ECG. 4. Mbraintrain smarting: This device is a mobile EEG with high time precision and good data quality. It is a wearable device that features body and head movements. 5. AliveCor Heart Monitor: It is a portable ECG recorder used to analyse the ECG recording and sends the data to medical professional for analysis. The data is accessed through a software app available in the device.

5 Challenges in Medical IOT The Wearable devices with respect to IoT platforms must offer simple, powerful application access. For medical parameter measurement there have been many platforms proposed by scientific, however there are stern challenges along these lines. Below are the four key constraints must be taken into considerations: 1. Secured connection and easy connection: A IoT system can do three level device administrations for a good platform, collecting data, storing the data permanently and observing the data in medical station after transmitting data to a hub. And this device administration must be secured. End to end security is provided by using data encryption method. All the device connections must be as simple as possible. 2. Power consumption: A device is said to be good if it consumes less power. Power loss is an important constraint to monitor the device for long term without any disruption. This constraint depends on many parameters such as framing the data with good frame format and size, writing an efficient and optimized c code and compression and encryption of the data. 3. Wearability: To monitor the patient’s health, specially the elder one, a sensor with wearable device is very helpful. The device should not be too heavy or bigger one. The device has to be simple and most importantly it must be comfortable to wear, so that elderly patients can carry and wear them easily. And it should be long-lasting. 4. Minimal data loss risk: While transferring data from microcontroller to cloud system or a mobile (smart phone), there are high chances of losing the data because of disconnection. This risk is reduced by transferring the data in frames and saving the data frame into buffers of microcontroller memory.

6 Conclusion Wearable devices are trending in various fields from sports, fitness to health monitoring. In particular, due to the increase in population throughout the world, wearable devices are becoming important for long-term health monitoring, especially

An Update on Effective Patient Monitoring …

213

for elderly people. The main intend of this chapter was to give an overview of this upcoming area of research and to report the full range of tools in health monitoring using wearable devices. In this chapter a brief description on the currently available technology is also discussed. All the parameters yielding to motion trackers, monitoring of vital signs are significant elements in health monitoring, a research on all these sectors were studied. In each field, a variety of methodologies are employed, but not all are efficient and effective. The most important criteria in this study were the possibility of using the device in the real world, performance, efficiency, and power consumption. In addition, we considered the price of each device. Finally, the most challenging bottleneck and some conclusions regarding the promising future in the IoT are presented.

References 1. Almeida, V.G., Nabney, I.T.: Early warnings of heart rate deterioration. In: International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 940–943 (2016) 2. Anzanpour, A., Azimi, I., Götzinger, M., Rahmani, A.M., TaheriNejad, N., Liljeberg, P., Jantsch, A., Dutt, N.: Self-awareness in remote health monitoring systems using wearable electronics. In: Design, Automation & Test in Europe Conference & Exhibition (DATE), pp. 1056–1061. IEEE (2017) 3. TaheriNejad, N.: Comprehensive observation and its role in self-awareness; an emotion recognition system example. In: FedCSIS Position Papers (2016) 4. Holtzman, M., Arcelus, A., Goubran, R., Knoefel, F.: Highly survivable bed pressure mat remote patient monitoring system for mHealth. In: Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pp. 268–271 (2012) 5. Eom, J., Jaisutti, R., Lee, H., Lee, W., Heo, J.S., Lee, J.Y., Park, S.K., Kim, Y.H.: Highly sensitive textile strain sensors and wireless user-interface devices using all-polymeric conducting fibers. ACS Appl. Mater. Interfaces 10190–10197 (2017) 6. Ge, G., Huang, W., Shao, J., Dong, X.: Recent progress of flexible and wearable strain sensors for human-motion monitoring. J. Semicond. (2018) 7. Waller, B.M., Cray, J.J., Burrows, A.M.: Selection for universal facial emotion. Emotion 435–439 (2008) 8. Liu, Y., Wang, H., Zhao, W., Zhang, M., Qin, H., Xie, Y.: Flexible, stretchable sensors for wearable health monitoring: sensing mechanisms, materials, fabrication strategies and features. Sensors 18, 645 (2018) 9. Yu, Y., Zhang, J., Liu, J.: Biomedical implementation of liquid metal ink as drawable ecg electrode and skin circuit. PLoS ONE 8(e58771), 83 (2013) 10. Patil, S., Pardeshi, S.: Health monitoring system using IoT. Int. Res. J. Eng. Technol. (IRJET) (2018) 11. Yamamoto, D., Nakata, S., Kanao, K., Arie, T., Akita, S., Takei, K.: All-printed, planartype multi-functional wearable flexible patch integrated with acceleration, temperature, and ECG sensors. In: IEEE 30th International Conference on Micro Electro Mechanical Systems, pp. 239–242 (2017) 12. Fuicu, S., Avramescu, A., Lascu, D., Marcu, M., Padurariu, R.: Realtime e-health system for continuous care. In: Proceedings of the 8th International Conference on Pervasive Computing Technologies for Healthcare, pp. 436–439 (2014) 13. Hameed, R.T., Mohamad, O.A., Tapus, N.: Health monitoring system based on wearable sensors and cloud platform. In: 20th International Conference on System Theory, Control and Computing (ICSTCC), pp. 13–15 (2016)

214

S. Rajendran et al.

14. Hameed, R.T., Mohamad, O.A., Hamid, O.T., Tapus, N.: Patient monitoring system based on e-health sensors and web services. In: 2016 Proceedings of the 8th International Conference on Electronics, Computers and Artificial Intelligence ECAI’16. (2016). ISSN: 1843-2115 15. Monicka, S., Suganya, C., Nithya Bharathi, S., Sindhu, A.P.: A ubiquitous based system for health care monitoring. Int. J. Sci. Res. Eng. Technol. (IJSRET) (2014). ISSN: 2278-0882 16. Meciak, V., Blaho, M., Mrafko, L., Mudrakova, T.: Sensor-based platform e-health connection with Matlab. Faculty of Electrical Engineering and Information Technology, Slovak University of Technology 17. Rakay, R., Visnovsky, M., Galajdova, A., Simsik, D.: Testing properties of e-health system based on Arduino. J. Autom. Control 122–126 (2015) 18. Hameed, R.T., Mohamad, O.A., Tapus, N.: Health monitoring system based on wearable sensors and cloud platform. In: 20th International Conference on System Theory, Control and Computing (ICSTCC), Sinaia, Romania, 13–15 Oct 2016

Automation of Assets Estimation for Accurate Tax Optimization Through Web Based Spatial Data Mining Techniques V. Pattabiraman and R. Parvathi

Abstract In the recent era, there is a growth of the huge volume of data; this includes the extension of spatial objects too. Hence, it is required to extract the knowledge from existing data sets and transform it into a human-understandable structure for further use. Spatial data mining is used to find the patterns in data with respect to geography. The calculation of income tax involves movable (e.g., two and four wheelers) and non-movable (e.g., houses, land etc.,) assets. Estimation of individual’s asset manually is difficult, because the information’s are distributed. Keywords Spatial data mining · Tax calculation and optimization · Spatial locality · Assets estimation · GIS

1 Introduction Spatial information is ordinarily put away as facilitates, topology and mapped information. It is regularly gotten to, control or analyze through Geographic Information System (GIS). GIS is the combination of cartography, measurable investigation and database innovation. The Geo spatial examination is an approach for applying factual examination and other enlightening procedures to topographically based information. Geographic Information System (GIS) could be a strategy utilized to capture, store, control, analyze, oversee and show all sorts of geologically alluded information. Controlling tax avoidance is one of the basic e-governance activities. This chapter is set to discover the conceivable outcomes of a person having disseminated and undeclared resources in a specific territory. The distinguishing proof of conceivable and anticipated resources for a person is done by localizing the zone of residence. A modern land documentation, registration framework is proposed to over-come all shortages of the existing one utilizing the spatial information mining. An inteV. Pattabiraman · R. Parvathi (B) School of Computing Science and Engineering, Vellore Institute of Technology, Chennai Campus, Vellore, India e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_14

215

216

V. Pattabiraman and R. Parvathi

gration of GIS procedure with web based spatial searching and distinguishing proof framework for resource observing will create parcels of convenience, in terms of data fundamental for online resource or property assess calculation and charging. The calculation of income tax includes movable (e.g., two and four wheelers) and non-movable (e.g., Houses, Land etc.,) resources. Estimation of individuals’ resource physically is troublesome, since the information’s are disseminated. By using Globalized Personal Identification Number (GPI) for registration, the government can effortlessly recognize the subtle elements of the resources in the topographically conveyed region. It moreover gives the points of interest of property, status of the claim properties. As of now, there is no arrangement to discover out the assess dodgers. Additionally the current system does not have all inclusive observing for land registration, which could be a drawback for the income tax department. Calculation of Spatial Adjoining connection between the spatial object. Spatial distance between spatial objects, arrangement of the goodness matrix, constraint-based spatial clustering are the basic procedures can be utilized to recognize the spatially related city information basically clinics, houses, schools, shops, transport stops, parks so on and so forward. The prior clustering handle cannot be done by spatial position, but it segments the spatial objects into a few parts. This chapter begins with the e-governance application system which is valuable in distinguishing the bunches of houses according to their sorts, values and geographic advertisement areas. With the assistance of this work the enterprises, metropolitan organizations, urban and country organization can issue the house tax evaluation arrange to the user online. Geo spatial analysis is an approach for applying measurable investigation and other enlightening methods for geologically based information. Geographic Information System (GIS) may be a strategy utilized to capture, store, control, analyze, oversee and show all sorts of topographically alluded information. A GIS will have geo spatial referencing system which uses the coordinates of latitude and/or longitude and at that point include additional data. Imperatively that data are recognized utilizing the same geo spatial referencing.

1.1 Benefits of Geographical Information System 1.1.1

Mapping Data

The central work of a geographic data framework is to provide a visual representation of information. It is assessed that 80% of the information we consider incorporating a geospatial component of a few frame. GIS provides a implies for that information to be put away in a database and at that point spoken to outwardly in a mapped format. Essentially, understanding the origin of objects and its the relationships within the spatial designs. From the illustration underneath, it is clear that the geographic highlights helps in understanding the difference between the wells and lakes.

Automation of Assets Estimation for Accurate Tax Optimization …

1.1.2

217

Proximity Analysis

This is an analytical method mainly utilized to define the association of a particular area with other areas and other focuses which are connected within a few ways. Numerous commercial associations utilize this analysis for recognizing locales applicable to trade outlets. This procedure considers distinctive variables like financial, social socioeconomics and nearness of contender outlets. The different subjects being included should utilize the same referencing framework for an exact proximity examination.

1.1.3

Buffering

Demonstrating the circle of impact of a point is carried out by utilizing a procedure called buffering that includes vicinity examination. A circle surrounding a particular point, line, or polygon (region) of an indicated remove is called buffering. Buffering is valuable for making a zone around a given geographic highlight for further analysis utilizing the overlay strategy. For illustration, a 1000 buffer can be created around a school to at that point utilize overlay examination to discover out how many libraries are inside 1000 of that school.

1.1.4

Find Clusters

Utilizing different calculations, it is conceivable to choose a gather of unrelated focuses on a topic that coordinates a set of criteria. A cluster seems to incorporate individuals where separate between them is less than a particular amount or zones where there’s the thickness of focuses more noteworthy than a particular level. Regularly a GIS will require different levels of cycle some time recently the proper calculations are distinguished.

1.1.5

Find Nearest

A strategy is used to find the degree of separations between the edge of a component and a point which may be characterized as a polygon utilizing vector focuses. From 1980s closest neighbor calculations has been a point of strongly investigated and unused approaches. This was characterized by scholastics like Benezecri and Juan in 1982. This calculation characterized centers on distinguishing focuses on the information set which are maximal or minimal or median individuals.

218

1.1.6

V. Pattabiraman and R. Parvathi

Location Analysis

The method best utilized to recognize an area for an unused retail outlet. The procedure has been created from hypothetical strategies utilized to clarify watched conditions to a calculation for distinguishing ideal areas. The calculations utilized tend to center on either maximal, negligible or middle individuals of a given dataset.

1.2 Application of GIS 1.2.1

Urban Arranging

GIS innovation is utilized to analyze the urban development and its heading of extension, and to discover reasonable destinations for advance urban advancement. In arranging to recognize the destinations, reasonable for the urban development, certain components have to be considered which is: Land ought to have legitimate availability, Land ought to be more or less level, Land ought to be empty or having low utilization esteem eventually and it ought to have a great supply of water.

1.2.2

Agricultural Applications

GIS can be utilized to make more successful and effective cultivating strategies. It can analyze soil information and to decide: what are the finest trims to plant?, where they ought to go? How to preserve nourishment levels to best advantage edit to plant? It is completely coordinated and broadly acknowledged for making a difference government agency to oversee programs that bolster ranchers and secure the environment. This could increment nourishment generation in numerous parts of the world so the world nourishment emergency may be avoided.

1.2.3

Landslide Risk Zonation Using GIS

Landslide danger zonation is the method of positioning distinctive parts of a zone concurring to the degrees of genuine or potential danger from avalanches. The evaluation of avalanche danger may be a complex assignment. It has ended up conceivable to productively collect, control and coordinated a assortment of spatial information such as geographical, auxiliary, surface cover and incline characteristics of a region, which can be utilized for danger zonation. The whole over said layer can well coordinate using GIS and weighted investigation is additionally accommodating to find Landslide inclined zone. By the assistance of GIS able to do hazard evaluation and can decrease the misfortunes of life and property.

Automation of Assets Estimation for Accurate Tax Optimization …

1.2.4

219

Surveying

Surveying is the estimation of area of objects on the earth’s surfaces. Land overview is measured the separate and points between distinctive focuses on the soil surface. An expanding, many nations, governments and territorial organizations are utilizing Global Navigation Satellite System (GNSS) estimations. GNSS is utilized for topographic studies where a centimeter level exactness is given. This information can be gained within the GIS framework. GIS apparatuses can be used to assess zone additionally, advanced maps can arranged.

2 Web-Based Geographic Information System (WebGIS) Technological propels within the Internet/Web have activated a move in the direction of Web-based geographic data framework (WebGIS). WebGIS points at giving GIS capacities (e.g., web mapping and spatial investigation) to clients through a familiar web browser, such as Web Pilgrim and Firefox. XML (Extensible Markup Language)/GML (Geography Markup Language)/SVG (Scalable Vector Graphics) based arrangements have been demonstrated to be a promising approach for building WebGIS [1–4]. In these arrangements, GML is utilized as a coding, putting away and transmitting benchmark of spatial information on the server side, whereas SVG is measured as a rendering device for showing spatial information on the browser side. As of late, as Google published a JavaScript library3 to supply “native” SVG bolsters on numerous browsers (e.g., Web Pioneer, Firefox, and Safari), XML/GML/SVG based WebGIS has ended up increasing. Despite, numerous of the XML/GML/SVG based WebGISs have been planned for visualization (web mapping) as it were, but “avoid they get to spatial investigative functions” e.g., Carto:net4, [5, 6]. To meet the expanding request of complicated GIS applications within the Web environment, spatial examination ought to be presented into XML/GML/SVG based WebGIS. There are moreover numerous XML/GML/SVG based WebGISs adopting a server-side arrangement to supply spatial investigation, i.e., executing all the spatial expository errands on GML on the server side, and sending the comes about to the browser side for visualization in SVG. This server-side arrangement gets to be illogical, as servers may not be able to handle a large amount of coexisting clients. Also, spatial analysis could be a complex assignment; clients regularly got to attempt distinctive inquiry arrangements some time recently they are fulfilled with the results. As spatial questions regularly result in a huge sum of information, there shall be a high transmission stack between the server side and the internet browser side. What’s more, in organizing to progress execution, not all the spatial questions ought to be executed on the server side. For illustration, the “Buffer” operation frequently comes about in more information yield than input. As the processor execution of a typical PC (individual computer) has been moved forward radically, “Buffer” may be more appropriate to be actualized on the browser side to decrease the network transmission stack.

220

V. Pattabiraman and R. Parvathi

Fig. 1 Content and steps of spatial data modeling

A promising arrangement for the other issues is to supply load adjusting for spatial examination between server and browser side. Stack adjusting spatial investigation executes spatial questioning and investigation either on the server side (GML) or on the browser side (SVG) depending on execution costs (i.e., organize transmission costs and computation costs). In arranging to actualize this stack adjusting spatial examination, GML-based spatial investigation (for the server side) and SVG-based spatial examination (for the browser side) ought to be created. There are a few investigates centering on spatial questioning on GML [7]. Despite, to the leading of our information, none of the investigate gives spatial querying and examination straightforwardly on SVG. Spatial information modeling is the method of abstracting the genuine world (recognizing the significant objects and wonders) and representing it in a suitable frame which can be recognized by computers [8]. SVG is one of the shapes which computers utilize to speak to important objects or marvels of the real world. In this manner, we will utilize the hypothesis of spatial information modeling, to talk about SVG-based spatial data representation to speak to these sorts of data. Three models play a part in spatial information modeling Fig. 1: conceptual, logical, and physical model. Agreeing to the substance and prerequisites of these models, spatial information modeling incorporates three steps [8]. (1) select a conceptual show, which can unique the genuine world most fittingly, (2) select a reasonable information structure to speak to the conceptual model, (3) plan a record arrange, or an appropriate strategy to record or store the information structure from step 2.

2.1 Spatial Conceptual Data Model Spatial conceptual information models can be categorized into the raster information show and vector information show. The last mentioned treats the world as a surface littered with recognizable spatial objects (e.g., cities, waterways), which exist autonomously of their areas [9]. As SVG is created to speak to vector design, this paper centers on the vector information demonstrate. Table 1 shows the comparison of raster and vector data models. Within the vector information demonstrate, spatial

Automation of Assets Estimation for Accurate Tax Optimization …

221

Table 1 A comparison of raster and vector data models Characteristic

Raster

Vector

Data structure

Usually simple

Usually complex

Storage requirements

Large for most data sets without compression

Small for most data sets

Coordinate conversion

May be slow due to data volumes, and may require resampling

Simple

Analysis

Easy for continuous data, simple for many layer combinations

Preferred for network analyses, many other spatial operations more complex

Positional precision

Floor set by cell size

Limited only by quality of positional measurements

Accessibility

Easy to modify or program, due to simple data structure

Often complex

Display and output

Good for images, but discrete features may show “stairstep” edges

Map-like, with continuous curves, poor for images

information are organized as a various leveled structure: spatial substance (question), layer, and an outline. A spatial substance alludes to a question or phenomenon with a geometrical shape. It has two sorts of traits: spatial qualities that portray geometry and topology of the spatial substance, and non-spatial properties which characterize the semantics (title, topic, etc.) of the spatial substance. Spatial substances having a place in the same subject frequently have comparative non-spatial properties or geometrical sorts, and in this way are assembled into a layer.

2.2 Spatial Data Structure In this section, a spatial data structure is planned to represent the over conceptual show based on object-oriented plan (OOD). Figure 2 depicts the spatial data structure. In this information structure, a spatial substance is planned as an abstract course Feature. The non-spatial traits of spatial substance are outlined as information individuals of the course Feature. Point, Bend, Surface, Multipoint, Multicurve, Multisurface and Multigeometry are acquired from the lesson Highlight. Outline is depicted as the course Outline, which incorporates information individuals such as x1, y1, width and stature (the bounding of the outline) and SRS (Spatial Reference System).

222

V. Pattabiraman and R. Parvathi

Fig. 2 Spatial data structure

2.3 Spatial Operators Spatial administrators are primarily outlined to get to spatial attributes, calculate spatial connections, and perform geometrical operations. The taking after administrators are outlined: (1) Attribute access operators: They incorporate Centroid, Length, Region and Envelope. They are utilized to calculate centroid, length, region and bounding of a spatial protest. (2) Spatial topological operators: Spatial topological relationships are exceptionally critical for spatial questioning and spatial examination. Two diverse approaches are utilized in portraying spatial topological relationships: DE-9IM (the Dimensionally Expanded 9 Crossing point Show) and RCC-8 (Locale Association Calculus). Agreeing to [10], these two totally distinctive approaches lead to an indistinguishable set of topological relations. This section employments the littlest total set of topological connections based on DE-9IM (Disjoint, Touch, Crosses, Inside and Cover), and executes them as our spatial topological administrators. For comfort, the Contain administrator is additionally included as the inverse of the Inside administrator. (3) Spatial metric administrators: They incorporate a Remove administrator to calculate the separate between two spatial objects. (4) Geometrical operators: Sometimes, spatial investigation should make modern spatial highlights with a few geometrical operations. In arrange to bolster this kind of operation, Crossing point, Union, and Distinction is characterized as geometrical administrators. A Buffer administrator is planned to support buffer area around a spatial protest. These five sorts of administrators meet the fundamental prerequisites of spatial examination. For arrange an investigation, Touch administrator and Length administrator can be utilized to discover out the touched spatial question (e.g., street) and the separate.

Automation of Assets Estimation for Accurate Tax Optimization …

223

3 Introduction to GML (Geography Markup Language) A mark-up language based on XML. The GML pattern definition is developed by the Open Geospatial Consortium (OGC) and gets to be an ISO standard. GML can be utilized to characterize spatial objects (highlights) with their geometry, qualities and relations and is aiming to be an open arrange for the trade of geospatial highlights between systems. 1. An XML language structure characterized by OGC to precise geological highlights. To assist clients and designers to structure and encourage the creation of GML-based application, GML gives GML profiles that are XML mappings that amplify the exceptionally GML specification in a secluded mold. A GML profile could be a GMLsubset for a concrete setting or application, but without the required for the total GML linguistic use, simplifying hence the selection of GML and encouraging its quick utilization. A few common cases of GML profiles that have been distributed are Point Profile, for applications with point geometric information, and GML Straightforward Highlights Profile, supporting vector includes re-quests and reactions, as within the case of a WFS. 2. Is the XML language structure characterized by the Open Geospatial Consortium (OGC) to specific geographic highlights? Geology Markup language is used in the Internet as a modeling language for geographical frameworks and also serves as an open compatibility organize for geographic exchanges. Conceptual plan may be a definitive step in the effective implementation of geographic applications. On the off chance that done accurately, the conceptual plan permits the early discovery of flaws, the satisfactory determination of representation options, and a vision of the framework necessities as to the foremost vital exchanges. There’s an agreement around the require for a correct conceptual plan as an early step within the plan of database-centric applications, and geographic applications are no exemption. Considering that require, a few conceptual models for geographic applications have been proposed. The Objection Modeling Technique for Geographic applications (OMTG) [1] is one of such models. It expands Unified Modeling Language (UML) concepts and charts to incorporate geographic representations and extraordinary sorts of connections. It moreover gives apparatuses for the plan of changes over the fundamental information representation and indicates visualization necessities for geographic information. The OMT-G demonstrates has been developed by our bunch and is as of now utilized by GIS engineers in numerous legislative, mechanical and scholarly organizations in Brazil. On a distinctive point of view, XML is being broadly embraced as a standard language for information representation, trade and capacity. Its auto-descriptive structure and it’s printed and non-proprietary organize (which encourages the creation of reports both by people and by program) are among the reasons behind this broad appropriation. The presence of a few extra languages for supporting and controlling XML reports, such as XPath, XQuery, XML Construction, DTD and Unwind NG, make its utilize indeed more alluring for information management. The most critical XML variety devoted to geographic information and applications, the Geography

224

V. Pattabiraman and R. Parvathi

Markup Language (GML1), is a standard created and advanced by the Open Geospatial Consortium (OGC2), in conjunction with GML Pattern. GML Pattern may be a geographic expansion of XML Pattern, the pattern definition language standard for XML information. In many situations, clients are anticipated to straightforwardly encode information and mappings in GML and GML Pattern, including within the setup of WebGIS bundles. Information put away in geo-enabled DBMSs, such as Prophet or PostGIS, can be traded to or imported from XML or GML. GML records are moreover utilized within the determination of OGC Web administrations such as Web Map Service (WMS3), Web Feature Service (WFS4) and others, which are important resources for building up spatial information frameworks. However, given the normal complexity and the idiosyncrasies of geographic applications, making a database structure straightforwardly in GML Construction isn’t simple. Besides, the interaction between application de-signer and master client, an essential assignment in the database application plan, is much harder to realize. It would be more normal to plan employing a conceptual show, taking advantage of the visual nature of lesson charts and other visual apparatuses for the interaction with the specialized client, sometime recently attempting to encode the database structure in GML. OGC utilizes the Geography Markup Language (GML) to precise topographical highlights. It is the XML code utilized by the OGC. Geology Markup language is used in the Internet as a modeling language for geographical frameworks and also serves as an open compatibility organize for geographic exchanges. GML not only incorporates inclusions and sensory information, but also vector or discrete objects. The capacity to coordinate all shapes of geographic information is the highlight of GML. GML is feature-centric. Highlights are substances things that portray viewpoints of the genuine world from the viewpoint of a specific application community whether circumscribed by topography or function or both. GML vocabularies are made by communities of interest. These vocabularies are called GML Application Mappings. On the off chance that you see at such an Application Construction with the help of world objects like Buildings, Streets, Buoys, Route Helps, Aircraft Flight Ways, Vehicles and Railroad Switches. Each such object is characterized within the construction by posting its properties. GML application mappings can be the premise of benchmarks themselves such as S57GML, cityGML, geoRSS GML and AIXM, or they can be casual manifestations for as it were a really little community, which in this case is up to the community. GML application schemas ought NOT be befuddled with GML profiles. A GML profile may be a subset of GML, characterized ordinarily by the subset instrument (portion of the GML detail), comprising of the chosen component, property and sort affirmations and all subordinate components from the GML center mappings (the patterns characterized by the GML detail). Application mappings can be built on GML profiles. A few GML profiles are moreover determinations counting the GML Basic Highlights Profile, the Point Profile, the GML Profile for GMLJP2 and the GML Profile for GeoRSS. GML was created to back geographic demands and transactions and this utilization originates before the WCS created for this reason. When a client sends a ask for geographic information—e.g. “find all water wells inside this county”—there must be a way to precisely “water well”, “county” and the “geometric degree of the county”.

Automation of Assets Estimation for Accurate Tax Optimization …

225

GML is utilized for this reason. When the client needs to send a exchange such as, “change the shape of the Holmes Waterway to the taking after …” the client needs a way to precisely the river’s geometry; GML gives this mechanism. Overview of GML Schema GML indicates XML encodings of a number of the conceptual classes characterized within the ISO 19100 arrangement of Worldwide Measures and the OpenGIS Unique Determination in conformance with these guidelines and specifications. In a few cases the mapping from the conceptual classes to XML is clear, whereas in other cases the mapping is more complex. In expansion, GML gives XML encodings for concepts not however modeled within the ISO 19100 arrangement of Worldwide Benchmarks or the OpenGIS Unique Detail. Cases incorporate moving objects, basic perceptions or esteem objects. Extra conceptual classes comparing to these expansions are too indicated in Add D. The GML construction comprises the components (XML components, attributes, straightforward sorts, complex sorts, trait bunches, benches, etc.) that are depicted in this Worldwide Standard. 3. GML Pattern Feature A GML could include encoded utilizing GML. Cases incorporate a street, a stream, an individual, a vehicle, a regulatory range, or an occasion. The highlight pattern gives a system for the creation of GML highlights and highlight collections. AbstractFeatureType The essential include model is given by the gml: AbstractFeatureType, characterized within the construction as follows:

The substance show for gml:AbstractFeatureType includes two particular properties appropriate for geographic highlights to the substance show defined in gml: AbstractGMLType. The esteem of the gml:boundedBy property depicts an envelope that encases the complete highlight occurrence, and is basically valuable for supporting fast looking for highlights that happen in a specific location. The esteem of the gml:location property depicts the degree, position or relative area of the highlight. gml:location is censured as portion of the standard substance show of gml:AbstractFeatureType. This unique component serves as the head of a substitution gather which may contain any components whose substance show is inferred

226

V. Pattabiraman and R. Parvathi

from gml:AbstractFeatureType. This may be utilized as a variable within the development of substance models. gml:AbstractFeature may be thought of as anything that’s a GML highlight and may be utilized to characterize factors or layouts in which the esteem of a GML property is “any feature”. This happens in particular in a GML include collection where the highlight part properties contain one or different duplicates of gml:AbstractFeature respectively. The other highlights are bounded By, BoundingShapeType, EnvelopeWithTimePeriod, EnvelopeWithTimePeriod-Type, locationName, locationReference, FeaturePropertyType, FeatureArrayPropertyType.

4 Asset Assessment Method Using WebGIS This section would empower to imagine the precise area of an individual’s resources in any portion in India. Within the existing framework for Land enlistment there’s no one of a kind recognizable proof for buyer or vender. By utilizing Globalized Individual Distinguishing proof Number (GPI) for Land enrollment, the government can effortlessly recognize the subtle elements of the resources in the geologically conveyed zone. Spatial information is ordinarily put away as facilitates, topology and mapped information. It is regularly gotten to, controlled or analyzed through Geographic Data Framework (GIS). GIS is the combination of cartography, measurable examination and database innovation. An unused Land documentation enrollment framework is proposed to overcome all setbacks of the existing one utilizing the spatial information mining. An integration of GIS strategy with web based spatial looking and recognizable proof framework for resource observing will create part of comfort, in terms of data basic for online resource or property charge calculation and charging. Also in case anybody needs to purchase a properties or the same individual needs to know the status of it’s possess properties, they can effectively retrieve the data from the proposed framework. At display there’s no arrangement to discover out the assess dodgers. Additionally the show framework does not have all inclusive checking for land registration which could be a impediment for the wage assesses office. By executing this proposed framework, government can categorize individuals concurring to the esteem of individual’s resources.

4.1 Definition of the Problem In the existing framework, Land enrollment is drained distributed way without any recognizable proof number. Each person may have undaunted properties like houses or Land anyplace in Tamil Nadu or India. But the tax appraisal may not be totally done. Computerized Land archive enlistment strategy is surrounded by integrating the information recovered from the income division into Google outline. This integration is done through Globalized Individual Recognizable proof Number (GPI).

Automation of Assets Estimation for Accurate Tax Optimization …

227

For example, in case client needs the points of interest from diverse cities, it can be effortlessly recovered from the Google outline based on Land marks. Consequently, there may be chances of assess avoidance. Within the proposed strategy, data of an individual’s like both mobile and undaunted resources can be recovered with offer assistance of this work.

4.2 Importance of the Proposed Project in the Context of Current Status Currently, there is no provision to assess the assets of an individual towards the tax calculation effectively. Assessment of tax is not carried out with the help of any unique identification of the citizen. Tax revenue to Government is purely based on the occupation of the individuals. If the probability of a tax inspection is increased, then automatically employers cannot evade from Tax payment. • Tax property is a vital and origin of revenue for public administrations. Based on the findings and by the methodology adapted to the hypothesis framed by the proposed technique, there is strong evidence for the investment patterns of the persons. These patterns exhibit affinity of locality, and spatial communal uniformity. Government may change the guideline value of land/place time to time which leads to tax evasions for many tax payers and they may conceal their geographically distributed properties.

5 Literature Survey 5.1 International Status The policy of the Land registration and Taxation is observed. It is useful for land administrator, decision makers through land strategy management, multi land researchers in land ownership with the help of land and it values. Objective of these papers is to identify the key issues related to the implementation of spatial data infrastructure and focusing on land information delivery services with respect to non technical integration perspective. It makes all the people to register their land through computerized methods using spatial data infrastructure in land administration [11–13]. The prediction of the land values depending upon the locations. Because the property tax is an important source of revenue for cities, the tax fluctuations in housing values may have significant budgetary consequences for local government. They analyzed time to time varying of house price periodically changes in the house values depend on the locations. This work concludes that, if there is a change in house price, then it will affect the other sources of income to the government. The

228

V. Pattabiraman and R. Parvathi

guideline value is directly propositional to the tax assessment of the property changes in any one values will reflect on the government [14]. An agent based model for land marketing. This model is used to investigate, visualize and explore the effects of land taxes on coastal zones. The model to posses the land prices and urban land patterns from down to top via interactions of individual agents in a land market. It will investigate how economic incentives in a land market that may influence the spatial distribution of land prices and urban development’s [15]. Raising tax for specific occupation is observed. It presents about 3 attributes. They are Occupation, Reputation, Tax compliance. The motivation of work is effectiveness of raising tax revenue by targeting specific occupations, if the probability of a tax audit is increase than automatically employers cannot evade from revenue. The main relationship is based on audit rates and compliance. The purpose of this work is increasing tax revenue for targeting people like specific occupations [16]. The old method of tax administration is that can develop into criminals and spatial important. It leads to illegal behavior through frequent audits. But in recent years this traditional method is complete. Instead of enforcement on tax payers, it focuses on provider of services to tax payer citizens [17–19].

5.2 National Status E-governance is the commitment to utilize and enhance governmental services and relationships in order to encourage the fair and efficient delivery of services. With the help of new technologies the author improved, the government process through online. The author contributed three different methods, 1. Improving government process, 2. Connecting citizens, 3. Building external interactions [20].

5.3 Methodology The methodology is used in predictive models in assets details by using person details. Mathematical methods are used for obtaining the combinations of the number of immovable assets. This is used to make relevant prediction. The extracted information can be represented in various formats, so the above two equations are one among them. Most these mathematical procedures are used to find out interesting relationship between variables and patterns. The homogeneous equation method can be used to find out the relationship among the attributes. This work consists of a number of clusters and these are arranged and shown in Sequence and series methods. GML may be a markup dialect that’s utilized to encode information around realworld objects. In GML, these real world objects are called highlights, and these can have geometrical and non-geometrical properties. The components of GML are uti-

Automation of Assets Estimation for Accurate Tax Optimization …

229

lized to depict highlights conjointly the connections between them. Highlights and properties constitute the crude information, which GML stores with something called an ‘instance document’. Separated from occasion archives, a moment sort of GML encoding exists, which is called ‘GML schema’. GML patterns are ‘vocabularies’ that characterize the structure of GML information. Since it is inconceivable to portray all highlights and predict their utilization and properties a priori, two sorts of GML mappings exist ‘GML center schemas’ and ‘GML application schemas’. GML center mappings contain all the essential components for depicting the topographical data and give the system that can be utilized to build GML application mappings. GML application patterns, and their corresponding information occurrences, are continuously domain-specific and ought to be made and managed by experts. The capacity to form GML application mappings gives the client the opportunity to describe and demonstrate boundless sums of assorted geospatial information. The Program can peruse GML application schemas and decide which components are highlights, and at that point utilize this data when experiencing information occasions that adjusts to those application mappings, utilizing ‘abstract types’. Program employments these theoretical sorts decide the nature of the components within the information stream [14, 21]. GML information components can have metadata properties, which can be alternatively joined to the GML component sorts. The metadata data itself contains it’s possess application pattern (or a few GML pattern archives imported into that application construction), which is comparative in structure to a GML application construction for highlights [22]. In truth, XML construction can be thought of as metadata for an XML record, and in arrange to construct pertinent XML pattern for an object’s metadata, a few GML application construction archives may be imported into the object’s metadata application construction GML can be utilized to form both metadata for robotization (i.e. machine-readable, in arranging or case to be utilized by web-services) and, similarly, human clear metadata which can be utilized straightforwardly by an end-user [23]. Most significance is the reality that in case the relevant and foundation fabric of the metadata is given, the method of revelation of data resources can be computerized. As such, GML encoded metadata could be a key enabler of information and framework interoperability [24]. GML is as it was concerned with the representation of geographic information, content and does not indicate how GML data should be displayed. To speak to the geo-spatial highlights within the frame of maps, GML information got to be changed through XSLT (Extensible Style sheet Language Transformations) in arranging to ‘style’ the GML geological substance into one of the common graphical designs (e.g. SVG, VML, X3D), which indicate how each component within the GML information ought to be shown [14, 25]. Web-services which prepare ‘request messages’ (sent to the benefit) and ‘response messages’ (sent from the service) that contain the geographic information are called Geospatial Web-Services shown in Fig. 3. GML can be utilized to depict these web-services as well as to depict the geospatial parts of their messages. With GML, geospatial web-services and supporting advances are utilized to distribute GML-encoded data over the Geo-Web (which incorporates moreover SDIs) [14].

230

V. Pattabiraman and R. Parvathi

Fig. 3 The server-based solution for GML/SVG

Fig. 4 Server based Architecture of GML/SVG

Internet solutions of GML/SVG Client-Server computing demonstrate (Fig. 4) is the key to fruitful information sharing on the Internet/Intranet. Essentially, Web components to SVG record. Certain expansion capacities must be given by the XSLT Motor to prepare the complex calculations since a GML record encompasses a part of geometrical coordinates. The DOM is an Application Programming Interface for substantial HTML and well shaped XML archives. Integration of SVG GIS takes after the client-server models. There are two varieties to the fundamental Web GIS application: the clientside and the server-side application the steps in GML outline, making can promptly be performed on either the client side or the server side. The application of GML/SVG in web GIS requires huge transmissions of information between client and server levels, and request for progressed spatial explanatory capabilities. All these prerequisites

Automation of Assets Estimation for Accurate Tax Optimization …

231

indicate that a overwhelming burden will be put on the server level, so the three-tier demonstrate is essential to corrupt the generally performances.

6 Conclusion With help of the WebGIS ,local body administrations can easily identify the non tax payer of their own immovable assets. With the help of mathematical modeling and number of hypothesis testing, better results can be achieved. This model can be used by Income Tax department to identify the persons who have paid their income tax.

References 1. Feng, W., Peng, L., Zhang, H., Lv, G.: Integration of WebGIS system based on .NET platform and web service method. Comput. Eng. 21 (2004) 2. Chang, K.T.: Introduction to Geographic Information Systems, pp. 117–122. McGraw-Hill Higher Education, Boston (2006) 3. Li, R.M., Lu, H.P.: Intelligent traffic management command and dispatch system based on WebGIS. Comput. Eng. 33(21), 232–234 (2007) 4. Yang, K., Sun, L.P., Huang, Y.X., Yang, G.J., Wu, F., Hang, D.R., Zhou, X.N., et al.: A realtime platform for monitoring schistosomiasis transmission supported by Google Earth and a web-based geographical information system. Geospat. Health 6(2), 195–203 (2012) 5. Yuan, K., Qin, W., Dong, M., Liu, J., Liu, P., Zhang, Y., Yang, W., et al.: Combining spatial and temporal information to explore resting-state networks changes in abstinent heroin-dependent individuals. Neurosci. Lett. 475(1), 20–24 (2010) 6. Köbben, B.: RIMapperWMS: a web map service providing SVG maps with a built-in client. In: The European Information Society, pp 217–230. Springer, Berlin, Heidelberg (2007) 7. Zhu, F., Guan, J., Zhou, J., Zhou, S.: Storing and querying GML in object-relational databases. In: Proceedings of the 14th Annual ACM International Symposium on Advances in Geographic Information Systems, pp. 107–114. ACM, Nov 2006 8. Chen, J., Li, C., Li, Z., Gold, C.: A Voronoi-based 9-intersection model for spatial relations. Int. J. Geogr. Inf. Sci. 15(3), 201–220 (2001) 9. Shekhar, S., Lu, C.T., Tan, X., Chawla, S., Vatsavai, R.R.: A visualization tool for spatial data warehouses. Geograp. Data Mining and Knowl. Discov. 73 (2001) 10. Renz, J.: Qualitative spatial reasoning with topological information. Springer (2002) 11. Shariff, A.R.M., Hamzah, A.H., Mahmud, A.R., Yusof, N.M.Z.N., Ali, H.M.: Spatial data infrastructure for malaysian land adminidtration. Int. J. Spat. Data Infrastruct. Res. 6 (2011) 12. Bertotti, M.L., Modanese, G.: From microscopic taxation and redistribution models to macroscopic income distributions. Int. J. Phys. A Stat. Mach. Appl. 390, 3782–3793 (2011) 13. Otusanya, O.J.: The role of multinational companies in tax evasion and tax avoidance: the case of Nigeria. Int. J. Critical Persp. Account. 22, 316–332 (2011) 14. Doerner, W.M., Ihlanfeldt, K.R.: House prices and city revenues. Int. J. Region. Sci. Urban Econ. 41, 332–342 (2011) 15. Filatova, T., Voinov, A., Van Der Veen, A.: Land market mechanisms for preservation of space for costal ecosystems: an agent-based analysis. Int. J. Environ. Model. Softw. 26, 179–190 (2011) 16. Forest, A., Kirchler, E.: Targeting occupations with varying reputation to increase tax revenue. J. Socio-Econ. 39, 400–406 (2010)

232

V. Pattabiraman and R. Parvathi

17. Alm, J., Cherry, T., Jones, M., Mckee, M.: Taxpayer information assistance services and tax compliance behavior. J. Econ. Psychol. 31, 577–586 (2010) 18. Xuming, M.: Research on corporate tax compliance, corporate governance and tax policies in China: from the international studies on the undersheltering puzzle, Department of Construction Management and Real Estate, Shenzhen University, Shenzhen, China, IEEE (2011) 19. Garrido, N., Mittone, L.: Tax evasion behaviour using finite automata: experiments in Chile and Italy. Expert Syst. Appl. (2011) 20. Shukla, A., Singh, C.P., Sen, M., Uppa, D.: EeGovernance: Inertia + Evolution  Revolution. In: National Conference of Computing for Nation Development (2011) 21. Park, J., Sandhu, R., Cheng, Y.: A user-activity-centric framework for access control in online. Univ. Texas San Antonio IEEE Internet Comput. 1089–7801, 62–65 (2011) 22. Sanchez, F.P.: Competitivity groups on social network sites. Math. Comput. Model. 52, 1052–1057 (2010) 23. Boldrini, C., Conti, M., Delmastro, F., Passarella, A.: Context and social—aware middleware for opportunistic networks. J. Netw. Comput. Appl. 33, 525–541 (2010) 24. Rodrigues, J.J.P.C., Sabino, F.M.R., Zhou, L.: Enhancing e-learning experience with online social networks. Inst. Eng. Technol. 5(8), 1147–1154 (2011) 25. Liu, Y., Yuvan, Y., Xiao, D., Zhang, Y., Hu, J.: A point-set-based approximation for areal objects: a case study of representing localities. Comput. Environ. Urban Syst. 34, 28–39 (2010)

Computationally Efficient and Secure HVS Based Composite Fingerprinting Scheme Using SHS and ECDSA Vineet Mehan

Abstract Fingerprinting is an extension to the watermarking principle where each copy of the image is embedded with unique purchaser’s information. Secure fingerprinting aims to prevent efforts to identify the embedded contents. It offers an effective tracing contrivance and protects the legitimacy of digital data. Only an authenticated user can retrieve the implanted contents. In this chapter a secure Human Visual System (HVS) based fingerprinting scheme is proposed for embedding unique composite fingerprints to safeguard buyer’s certification accompanied by image integrity. Combination of authentication and reliability issues is determined using Secure Hash Standard (SHS): Secure Hash Algorithm (SHA-512/256) and Elliptic Curve Digital Signature Algorithm (ECDSA). SHS generates a unique authentication code by integrating image information with buyer’s credentials. Uniqueness property of digest allows for identification of authentication deceit by ensuring maximum collision resistance strength. Composite fingerprint (CF) created using ECDSA is embedded using a maximally separated imperceptible mode, by means of HVS approach. The positions take advantage of color content of digital images to reduce the visible distortion introduced by embedding CF. Signing and verifying digital signatures are two core elements adopted while embedding and retrieving of CF. Test cases are generated for different image dimensions used for embedding varying size fingerprints. Experimental results prove the approach to be beneficial by forming a maximally resilient counter to collusion attacks. An efficient imperceptible semi-blind fingerprinting scheme is achieved without degrading the quality and integrity of the host image. Keywords ECDSA · Fingerprinting · HVS · SHA · SHS · Watermarking

V. Mehan (B) Department of Computer Science & Engineering, Maharaja Agrasen University, Baddi 174103, HP, India e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Kumar and U. K. Wiil (eds.), Recent Advances in Computational Intelligence, Studies in Computational Intelligence 823, https://doi.org/10.1007/978-3-030-12500-4_15

233

234

V. Mehan

1 Introduction Protection of intellectual digital property raised the need of Digital Rights Management (DRM). Watermarking (W ) originated to safeguard the copyright of digital images with its original owner. W has emerged as a prominent practice in the last decade. Digital watermarking is a technique of implanting data into digital image [1]. Embedded data in digital image is termed as a watermark (W m ). Inserted W m can be retrieved for ensuring copyright credentials for verification purpose [2–5]. Fingerprints are an addition to the watermarking technique. In fingerprinting (F) distinctive buyer’s credentials are inserted into digital image [6]. The inimitable number so inserted is termed as fingerprint (F p ). For the proprietor, implanting an exclusive F p is a good way to perceive buyers who violate the license contract by stealing the secure data. The following are some requirements that a F p algorithm may possess: Fidelity: Fidelity is a significant property of all perceptual based fingerprinting. Fidelity identifies the observable likeness amongst cover image and F p image. Fingerprinting scheme has to be prepared in a manner such that it does not disturb the quality of the host image and of the concealed [7]. The changes in the image should not be noticeable by naked eye [8]. To encounter this restriction, the perceptual alteration observed due to implanting of W m is kept under the masking threshold of HVS [9]. Payload Capacity: It is very significant to find the maximum amount of information that can be safely hidden in an image [10]. Various applications have different sizes of the data to be hidden. This directly affects the robustness and the perceptual impact [11]. A considerable ratio is to be determined for maintaining appropriate imperceptibility and the amount of data to be embedded. Computational Cost: Cost of embedding F p is another significant criterion for fingerprinting schemes. Embedding cost is determined by the amount of time F algorithm takes while insertion of data [12]. Retrieval cost is determined by the amount of time F algorithm takes while retrieving of F p . The amount of time taken by the F scheme while embedding and retrieving should be less in order to make it more efficient. Security: Secure fingerprinting aims to prevent any effort by a rival to find out what are the contents of the embedded F p . Only an authenticated user can retrieve the implanted contents. Secure F p insertion technique can be applied by using symmetric ciphers, digital signatures [13] and encryption algorithms [14, 15]. In this chapter, we consider a secure fingerprinting system by incorporating latest hash algorithm and ECDSA. A small and protected F code for images is proposed by Kim and Suh in 2004 [16]. Fingerprint hinges on the buyer’s uniqueness. Collusion attacks are prevented in our proposed approach by varying the size of the F p code in quotient to the number of customers. A digital W scheme based on composite watermarking is proposed by Chang et al. in 2005 [17]. In this approach a new robustness improvement retrieval technique is presented by overlaying a number of

Computationally Efficient and Secure HVS …

235

weighted copies of condensed size W m . Composite W m is generated by decreasing the original W m and then replicating it t times. Generated composite W m is then encrypted and inserted into the original cover image. In our fingerprinting model, it is challenging for colluders to execute a collusion attack because of varying size of CF. Varying size CF’s are injected by adopting a maximally alienated imperceptible mode. Each bit sequence of CF varies from the other with a minimum bit length of 32 bits. The maximum variation between successive bits sequence can be 768 bits. The buyer credentials are added with each successive insertion of CF, and there can be b + 1 customer, when the numbers of buyers is b in a multi owner ship scenario. This chapter is organized as follows. Section 2 describes the previous work related to of the F technique and collusion attacks. Section 3 presents fingerprinting with Secure Hash Standard; attacks launched against hash algorithms which proved to be successful and NIST latest standards incorporating the collusion resistance strength. The ECDSA generated signature for signing and verification is evaluated in Sect. 4. Section 5 describes the performance of the proposed fingerprinting scheme in experiments using 100 images of 4 different resolutions. Quantitative parameter analysis results are depicted. Finally, conclusions are drawn in Sect. 6.

2 Previous Work In this section, we briefly review F scheme and related works. Let L be the length of a CF, and N be the number of customers in a F scheme. For the collusion attack at least bmax buyers must collude to produce a plagiarized duplicate. Lee and Yoon [18] set the length of F code to 64 bits above for expressing enough number of the user. The binary F p of the m-th user is indicated by k m,n ∈ {0, 1}, (1 ≤ n ≤ L), where k m,n is generated from an autonomously and identically scattered random number with a probability pi such that Pr [k m,n  1]  pi and Pr [k m,n  0]  1 − pi . This probability pi follows a continuous distribution over an open unit interval (0, 1), which is titled the bias distribution. System put forward an amalgamation between DRM and F. In this chapter a method for effective coordination is proposed that interconnect F scheme with DRM structure. In order to affirm sufficient number of the users F p length is set between 288 and 1056 bits. A group based F scheme is proposed by He and Wu [19]. Examination of prevailing group-based technique reveals that the detection precision is sensitive to the group detection threshold, and the threshold attaining upright routine is associated to the collusion pattern. Fingerprint sequence is constructed by tx y 

 √ 1 − ρ f ab + ρxa

(1)

where ρ is used to regulate the correlation between users in individual group; f ab contains distinct user evidence and x a are the spreading sequence of group informa-

236

V. Mehan

tion. An automatic adaptive detection technique is designed based on the detection statistics of group data to become accustomed with diverse collusion configurations. Correlation based method adopted for group detection is, (z − x)(ta1 + ta2 + · · · + ta M ) TG (a)   ||t||2 [M + (M2 − M)ρ]

(2)

where z is the colluded version, ||t|| is the strength of F p , (t a1 , …, t aM ) are the members of group. Experimental results reveal an enhancement of 10% on the complete detection probability. A 32 bit threshold level is set between successive fingerprints in our suggested system for refraining from collusion attack. The high bit threshold level correctly identifies the colluder. Swaminathan et al. [20] proposed a model for estimation of fingerprints. The paper classified fingerprints into two categories: intrinsic and extrinsic. Intrinsic fingerprints are generated using in-camera processing setups. Extrinsic fingerprints are inserted at the time of formation of multimedia data. Forensic signal processing procedures confirmed the presence and absence of intrinsic F p in host image. Similarity between the coefficients of the test input and the reference image using the similarity score defined as      θt (m, n) − μt ) × θ f p (m, n) − μ f p s θt , θ f p 

(3)

where μt denote mean of θ t and μfp represents mean of θ fp . Test image is then categorized as unmanipulated if the similarity to the F p pattern is greater than a suitable threshold. A good correspondence among targeted coefficients resembles no alterations and confirms the integrity of the host image. In our projected approach a composite message is created by integrating image information with buyer’s details. The problem of F compressed signals is determined by Varna et al. in 2009 [21]. The paper provides a solution to the collusion attack by designing collusionresistant fingerprints. F scheme inserts F p in each legitimately circulated copy of the image that distinctively ascertains the inheritor. Our proposed technique effectively blends the authentication and integrity concerns using enhanced hash algorithm with different truncation levels applied to CF for preventing collusion attack. One test case for the proposed approach also contains the CF j raised by adding ACD signal with Gaussian F p Qj to the quantized host image and then applying re-quantization.   C F j  r ound( S j + d j + Q j /Δ) × Δ

(4)

where d denotes the random variables uniformly distributed; Δ is the step size and S j are multiples of Δ. A buyer-seller protocol is designed by Rial et al. in 2010 [22]. The protocol is secret as the identity of the buyers (By ) is undisclosed. The protocol applies dual cryptographic primitives which include: group signatures (Gs ) and homomorphic encryption (HE). Gs permit By to sign the acquisition media on behalf of group of

Computationally Efficient and Secure HVS …

237

By . The system is dynamic as new members can be added to the group easily. HE lets By and seller (S l ) to cooperatively work on encrypted W m (W E ) to be inserted in the host image (HI ). To handle joint ownership of different owners Mohanpurkar and Joshi in 2011 [23] suggested a model to W . Technique is projected for copyright security and for ascertaining traitor credentials by F. The paper suggests designing of new algorithm for relational databases by exploiting W . W ascertains the source of data, stops illegal replication and circulation of digital content. A content-based image F mechanism is proposed by Lv and Wang in 2012 [24]. Fingerprint is created by hashing the image. Instead of embedding F p in the image a separate database is maintained for storing the image hash. In our proposed approach latest SHA is used to generate a secure message digest.

3 Fingerprinting with Secure Hash Standard Hash algorithm (HA) computes original message into a lesser size message abstract. Any alteration done to the message perpetually formulates a dissimilar hash output when similar HA is applied. In the proposed approach a composite message (CM) is created by integrating image information with buyer’s details (Bd ). CM contains Red color channel (Rc ) and Green color channel (Gc ) pixel information. Color image W [25] is suitable for computing the perceptual redundancy characteristic in color channels (C c ) of the cover color image. Blue color channel (Bc ) is selected for embedding the signature as HVS is extra sensitive to red color while least sensitive to blue color. A framework to embed color W m image to color host image in an effective way is given by Das et al. in 2009 [26]. An invisible blind technique inserts the W m image in the blue channel using 2-bit LSB scheme which was an improvement over one bit scheme. In the proposed F framework, 8-bit blue color channel is used to embed the F p . A few desirable properties of hashing function H k (.) include Noninvertible Function: I → Hk (I )

(5)

Compactness: Si ze (Hk (I ))  Si ze( I )

(6)

Visible Robustness: Pr (Hk (I )) ≈ (Hk (Id )) ≥ 1−ε, 0 ≤ ε < 1

(7)

(Hk (I’) ≥ 1 – г, 0 ≤ г < 1

(8)

Parameters ε and g should be close to zero. Bd include Buyer Identification Number (BIN), Buyer Name (BN), Buyer House Number (BHN), Buyer Sector (BSE), Buyer City (BC), Buyer State (BST ), Buyer Mobile Number (BMN), Buyer E-mail (BEM) and Buyer Country (BCY ). Information pertaining to buyer is stretched depending upon F p size generated after integra-

238

V. Mehan

tion of Bd with C c to form a CM. Different size F p aids in prevention of collusion attacks. SHA-512/256 standard acts as an intermediate step in the generation of digital signature. Collision resistance strength ensures unviability to find two dissimilar inputs to the H k that have the same hash value such that   hash(x)  hash x

(9)

where x and x are two different inputs. Preimage resistance strength specifies choosing a random hash value (H v ) such that H ash(x)  Hv

(10)

The second preimage resistance strength of SHA-512, is minimum of (L – M) and λ, where L is the output block size of the cryptographic hash function for any truncated message digest of λ bits; and M is a function of the input block size, as follows:   max_message_length_in_bits (11) M  log2 input_block_si ze_in_bits where max_message_length_in_bits ≥ input_block_size_in_bits. User with an appropriate HA can verify the authenticity, integrity and ownership of an image. If the user performs the W m extraction with an incorrect key and inappropriate hash function, the user obtains an image that resembles noise. NIST enumerated Security strengths for hash functions [27] is given in Table 1. Collision Resistance Strength (CRS) specifies the computation complexity to find out similar hash value from two dissimilar messages. Pre-image Resistance Strength (PRS) identifies the non-possibility to find original message from M d . Second Preimage Resistance Strength (SPRS) of hash algorithm aims at identification of another message from the hash value that gives the same M d as the original message. Literature investigation reveals the use of hash algorithms like MD5 and SHA-1 in W and F applications. To maintain legitimacy and reliability of digital mammography images Zhou et al. in 2001 [28] proposed an effective system. Similarly a digital W system for copyright security of images is given by Samuel and Penzhorn in 2004

Table 1 Strength of security properties of the hash algorithms Strength (in bits)

SHA-1

SHA-224

SHA-256

SHA-384

SHA-512

CRS