Skip to main content

Über dieses Buch

This book is a collection of peer-reviewed best-selected research papers presented at 4th International Conference on Computer Networks and Inventive Communication Technologies (ICCNCT 2021). The book covers new results in theory, methodology, and applications of computer networks and data communications. It includes original papers on computer networks, network protocols and wireless networks, data communication technologies, and network security. The proceedings of this conference are a valuable resource, dealing with both the important core and the specialized issues in the areas of next-generation wireless network design, control, and management, as well as in the areas of protection, assurance, and trust in information security practice. It is a reference for researchers, instructors, students, scientists, engineers, managers, and industry practitioners for advanced work in the area.



Energy Efficient Clustering in Wireless Sensor Networks by Opposition-Based Initialization Bat Algorithm

Bacanin, Nebojsa Arnaut, Uros Zivkovic, Miodrag Bezdan, Timea Rashid, Tarik A.Wireless sensor networks belong to the group of technologies that enabled emerging and fast developing of other novel technologies such as cloud computing, environmental and air pollution monitoring, and health applications. One important challenge that must be solved for any wireless sensor network is energy-efficient clustering, that is categorized as NP-hard problem. This led to a great number of novel clustering algorithms, that emerged with sole purpose to establish the proper balance in energy consumption between the sensors, and to enhance the efficiency and lifetime of the network itself. In this manuscript, a modified version of the bat algorithm, that belongs to a group of nature-inspired swarm intelligence metaheuristics, is proposed. Devised algorithm was utilized to tackle the energy-efficient clustering problems. Performance of proposed improved bat metaheuristics has been validated by conducting a comparative analysis with its original version, and also with other metaheuristics approaches that were tested for the same problem. Obtained results from conducted experiments suggest that the proposed method’s performance is superior, and that it could bring valuable results in the other domains of use as well.

Nebojsa Bacanin, Uros Arnaut, Miodrag Zivkovic, Timea Bezdan, Tarik A. Rashid

Efficient Data Collection in Wireless Sensor Network

There has been a steady increase in climatic and man-made disasters for the past couple of years. Due to this, numerous researches on Mobile Wireless Sensor Networks (MWSN’s) are attained, the sensors which are mobile and can do numerous tasks such as go to places which are dense and do not have access to the physical human body; compensation in cost, flexible, and many more which the ordinary Wireless Sensor Networks (WSN’s) lack. Many ideas have been used to give the best out of the MWSN’s. In every idea so far, the Mobile Sensor (MS) collects data from the sensor and return to the station which takes time, and this delays the data which was to be returned early in order to take actions, also the capacity of the storage in MS is quite low, so a number of laps are required to collect data from all sensors. In this paper, the Models proposed can eradicate such problems by using MS as a means of connectivity, which will connect sensors and the station.

Meet J. Vasani, Satish Maurya

Job Scheduling in Cloud Computing Based on DGPSO

In a Cloud Computing environment, dynamic and uncertain nature makes task scheduling problems more complex. It states the need for efficient task scheduling designed and implementation as a primary requirement for achieving QoS. A proper resource utilization enables maximum profit for the Cloud providers. The best scheduling algorithm does not consider the task set collected from the users, but it considers the resources provided by providers for operating the tasks. In this paper, we propose a Dynamic Group of Pair Scheduling and Optimization (DGPSO) algorithm. The proposed DGPSO is the performance-enhancing of AWSQP by using VM pair implementation and partition-based priority system into three levels. These three levels in the priority system such as low, medium, and high. According to the task size, the VM pairing is done. For this, the VM's parameters include communication time, system capacity, memory size, and processing speed. On the dataset, the task sizes are examined and separated according to the priority levels. On which the high priority comprises video files, the audio files under medium-level priority, and the remaining text documents, ppts, etc. included in under low priority levels. Based on the proposed task scheduling mechanism, an experiment is conducted on the aspects of computation cost, communication cost, execution time, CPU utilization, and bandwidth. The obtained results prove its achieved performance is far better than the existing approaches.

J. Arul Sindiya, R. Pushpalakshmi

Read–Write Decoupled Single-Ended 9T SRAM Cell for Low Power Embedded Applications

Static Random-Access Memory (SRAM) is the most significant building block of embedded Systems and microprocessor. Traditional 6T cell used as a data storage element in the SRAM cell but is suffered from low stability, low process tolerance and high-power consumption issue. Technology is continuously scaling down into the nanometer regime to achieve higher integration. Minimum size cell is used to achieve higher integration density in nm technology node but it significantly increases the leakage current and decreases stability. These issues are more critical in the conventional 6T cell. This article introduces a new read/write decouple single-ended 9T cell with high stability, low process tolerance, and low static and dynamic power consumption. This 9T cell shows higher read/write stability due to read buffer and dynamic loop cutting techniques respectively. Furthermore, it shows the low leakage current due to the stack transistor technique and low dynamic power due to a single bit line (BL). In contrast to the traditional 6T SRAM cell, the proposed 9T cell has a 4.28 × higher Read Static Noise Margin (RSNM), 1.06×  higher Write Static Noise Margin (WSNM), and approximately the same Hold Static Noise Margin (HSNM). The proposed 9T cell reported 0.48× lower power consumption compared to the conventional 6T cell. This 9T cell shows the half select free operation and aids bit interleaving architectures therefore it may be an appealing choice for low power embedded system.

Amit Singh Rajput, Arpan Dwivedi, Prashant Dwivedi, Deependra Singh Rajput, Manisha Pattanaik

Spam Detection Using Genetic Algorithm Optimized LSTM Model

Sinhmar, Abhinav Malhotra, Vinamra Yadav, Raghav Kumar, ManojThe advancement in technology over the years has resulted in the increased usage of SMS which in turn has provided certain groups a chance to exploit this service by spreading spam messages to consumers making it difficult for people to receive important information and also possessing a threat to their privacy. There are numerous machine learning and deep learning techniques that have been used for spam detection and have proved to be effective. But in deep learning techniques, it is essential to fine-tune the hyperparameters which requires excessive computational power and time, making the process less feasible. The proposed work aims at reducing this computational barrier and time by using Genetic Algorithm in order to select the key hyperparameters. A randomly generated population of LSTM models was created and further generations were produced following the different stages of the genetic algorithm multiple times until the terminal condition was met, and the performance of each candidate solution was evaluated using a chosen fitness function. The most optimal configuration was obtained from the final generation which is used to classify the messages. Four metrics, namely the accuracy, precision, recall and f1-score were used to analyze the model’s performance. The experimental results demonstrate that the Genetic Algorithm optimized LSTM model was able to outperform the other machine learning models.

Abhinav Sinhmar, Vinamra Malhotra, R. K. Yadav, Manoj Kumar

Affine Recurrence Based Key Scheduling Algorithm for the Advanced Encryption Standard

Encryption algorithms such as Advanced Encryption Standard (AES) are known as symmetric encryption algorithms which use the same key for both encryption and decryption. These algorithms have a huge variety of applications such as for securing data and transferring files. They also have a key expansion algorithm which is used for expanding the given key. In AES the key expansion algorithm expands the given 128 bit key into 176 bytes which will then be used in the encryption process which spans for 10 rounds. This paper aims at making this process even more secure. This is implemented by including different steps to increase security in this key expansion process so that it becomes computationally infeasible to get the original key even if the adversary gets a hold of the different parts of the key. Current works suggest that there is a need for increased security in the key scheduling algorithm of the AES. It has also been tested to have inferior strict avalanche criteria in comparison with other contenders of the AES such as Serpent and Twofish. Usage of the concept of affine recurrence ensures this in the proposed model. In affine recurrence, no two outputs of the operation will have a relation between them. Another concept used is the AES Substitution Box (S-box), and this is done to ensure higher levels of confusion in the key scheduling process.

S. Shashankh, Tavishi Kaushik, Svarnim Agarwal, C. R. Kavitha

Simplify Your Neural Networks: An Empirical Study on Cross-Project Defect Prediction

Malhotra, Ruchika Khan, Abuzar Ahmed Khera, AmritEnsuring software quality, when every industry depends on software, is of utmost importance. Software bugs can have grave consequences and thus identifying and fixing them becomes imperative for developers. Software defect prediction (SDP) focuses on identifying defect-prone areas so that testing resources can be allocated judiciously. Sourcing historical data is not easy, especially in the case of new software, and this is where cross-project defect prediction (CPDP) comes in. SDP, and specifically CPDP, have both attracted the attention of the research community. Simultaneously, the versatility of neural networks (NN) has pushed researchers to study the applications of NNs to defect prediction. In most research, the architecture of a NN is arrived at through trial-and-error. This requires effort, which can be reduced if there is a general idea about what kind of architecture works well in a particular field. In this paper, we tackle this problem by testing six different NNs on a dataset of twenty software from the PROMISE repository in a strict CPDP setting. We then compare the best architecture to three proposed methods for CPDP, which cover a wide range of scenarios. During our research, we found that the simplest NN with dropout layers (NN-SD) performed the best and was also statistically significantly better than the CPDP methods it was compared with. We used the area under the curve for receiver operating characteristics (AUC-ROC) as the performance metric, and for testing statistical significance, we use the Friedman chi-squared test and the Wilcoxon signed-rank test.

Ruchika Malhotra, Abuzar Ahmed Khan, Amrit Khera

Emotion Recognition During Social Interactions Using Peripheral Physiological Signals

Gupta, Priyansh Balaji, S Ashwin Jain, Sambhav Yadav, R KThis research aims to present a method for emotion recognition using the K-Emocon dataset (Park et al. in Sci Data 7(1):1–16 [8]) for use in the healthcare sector as well as to enhance computer–human interaction. In the following work, we use peripheral physiological signals to recognize emotion using classifier models with multidimensional emotion space models. These signals are collected using IoT-based wireless wearable devices. Emotions are measured in terms of arousal and valence by using physiological signals obtained from these devices. Several machine learning models were used for emotion recognition. Thirty-eight input features were extracted from a variety of physiological signals present in the dataset for analysis. Best accuracy achieved for valence and arousal in our experiment was 91.12% and 62.19%, respectively. This study targets recognition and classification of emotions during naturalistic conversations between people using peripheral physiological signals. It is shown that it is viable to recognize emotions using these signals.

Priyansh Gupta, S. Ashwin Balaji, Sambhav Jain, R. K. Yadav

Phishing Detection Using Computer Vision

Phishing is a cyber-crime wherein innocent web users are trapped into a counterfeit website, which is visually similar to its legitimate counterpart, but in reality, it is fake. Initially, users are redirected to phishing websites through various social and technical routing techniques. Users being ignorant about the illegitimacy of the website, provide their personal information such as user id, password, credit card details, bank account details to name a few. These details are stolen by the phishers and later used for either financial gains, or to tarnish a brand image or even more grave crimes like identity theft. Many phishing detection and prevention techniques are proposed in the literature; however, there is much scope in the cyber-security world with the advent of smart machine learning and deep learning methods. In this research, we explored computer vision techniques and build deep learning and machine learning classifiers to detect phishing website and their brands. Some of the experiments include Transfer Learning and Representation Learning techniques by utilizing various off-the-shelf Convolutional Neural Network (CNN) architectures to extract image features. It is observed that DenseNet201 outperforms all experiments conducted as well as the existing state-of-the-art on the dataset used, proving the hypothesis that Convolutional neural networks are an effective solution for extracting relevant features from phishing webpages for phishing detection classification.

Shekhar Khandelwal, Rik Das

A Comprehensive Attention-Based Model for Image Captioning

Kumar, Vinod Dahiya, Abhishek Saini, Geetanjali Sheokand, SahilImage captioning/automatic image annotation is referred to as description of image in text according to the contents and properties observed in a picture. It has numerous implementations such as its utilisation in virtual assistants for people with visual impairment, for social media and several other applications in computer vision and deep learning. Another interesting application is that a video can be explained frame by frame by image captioning (considering it to be carousel of images). In this paper, we have used an encoder–decoder architecture along with attention mechanism for captioning the images. We have used layers of CNN in the form of an encoder and that of RNN as decoder. We used Adam optimiser which gave the best results for our architecture. We have used Beam Search and Greedy Search for evaluating the captions. BLEU score was calculated to estimate the proximity of the generated captions to the real captions.

Vinod Kumar, Abhishek Dahiya, Geetanjali Saini, Sahil Sheokand

Multimedia Text Summary Generator for Visually Impaired

Banerjee, Shreya Sirigeri, Prerana Karennavar, Rachana B. Jayashree, R.With the advancing methodologies in the field of NLP, text summarization has become an important application and is still under research.Taking into consideration today’s busy world, time is a most important factor for everyone and people do not bother to make time to listen to long audio news or read long news articles, and as the visually impaired are an important part of our society, it becomes still difficult for them to read although there is braille, but it is inconvenient for them every time to read through braille. The aim is to make use of advancing technology to make their lives easier. This gave us inspiration that led to the idea of generating concise and short summaries for the visually impaired. Our system makes use of various APIs like speechrecognition, pyspeech, Google Cloud Speech API, etc., to extract text and then use summarisation techniques to present the most accurate summary and convert it back to audio format so that the news is more accessible for the visually impaired.

Shreya Banerjee, Prerana Sirigeri, Rachana B. Karennavar, R. Jayashree

Keylogger Threat to the Android Mobile Banking Applications

Android is presently the world’s most prevalent operating system, reaching more mobile customers than any other operating system to date by providing numerous services via smartphone and various android devices to make our life easy. Most of the android applications are developed by third-party android developers, android provides them an enormous platform to build their application. Modern cyber attackers are highly interested in this platform to access user’s sensitive information; with their own build malicious application or take amenities of other android developer’s application to spy on user’s activity. We have found that keyloggers can thieve personal information from users, such as credit card information or login pin/password from their typed keystroke in social networking and mobile banking apps. In case of mobile banking generally the mobile devices such as smartphones, tablets are being used for financial communications with the banks or financial institutions, by allowing clients and users to conduct a variety of transactions. In android app store (Google Play) keylogger apps are initially blocked but using some vulnerabilities in app permission it can be installed with benign and trusted apps. Both expert and maladroit android smartphone users use the mobile banking application, inexpert users are unable to find the vulnerabilities and attacker’s use this as an advantage to place an attack. The security android has provided for all the application is not sufficient for the sensitive application such as mobile banking application. In our paper, we discuss how attackers steal mobile banking app users sensitive information for their financial gain and also proposed a method to avoid keylogger attacks on android mobile banking apps.

Naziour Rahaman, Salauddin Rubel, Ahmed Al Marouf

Machine Learning-Based Network Intrusion Detection System

As the network is dramatically extended, security has become a significant issue. Various attacks like DoS, R2L, U2R are significantly increasing to affect these networks. Thus, detecting such intrusions or attacks is a major concern. Intrusions are the activities that breach the system's security policy. The paper's objective is to detect malicious network traffic using machine learning techniques by developing an intrusion detection system in order to provide a more secure network. This paper intends to highlight the performance comparison of various machine learning algorithms like SVM, K-Means Clustering, KNN, Decision tree, Logistic Regression, and Random Forest for the detection of malicious attacks based on their detection accuracies and precision score. A detailed analysis of the network traffic features and the experimental results reveal that Logistic Regression provides the most accurate results.

Sumedha Seniaray, Rajni Jindal

BGCNN: A Computer Vision Approach to Recognize of Yellow Mosaic Disease for Black Gram

The yellow mosaic disease is a common black gram leaf disease that causes severe economic losses to local farmers and a hindrance to healthy production which can be prevented by computer vision based fast and accurate recognition system. In this paper, Black Gram Convolutional Neural Network (BGCNN) has been proposed for the recognition of this disease, and the performance of BGCNN has compared with the state-of-the-art deep learning models such as AlexNet, VGG16, and Inception V3. All the models have trained with original dataset having 2830 images and expanded dataset generated with image augmentation having 16,980 images that increase test accuracy of all the models significantly. BGCNN realizes accuracy of 82.67% and 97.11% for the original and expanded dataset, respectively. While, AlexNet, VGG16, and Inception V3 have achieved 93.78%, 95.49%, and 96.67% accuracy for the expanded dataset, respectively. The obtained results validate that BGCNN can recognize yellow mosaic disease efficiently.

Rashidul Hasan Hridoy, Aniruddha Rakshit

Irrelevant Racist Tweets Identification Using Data Mining Techniques

In recent times, Twitter is one of the major sources to access information. Its feature of the hashtag is something that grabs more attention from the users. One can write one’s mind and heart out on Twitter at any given minute. Due to which there is a rapid increase in the generation of irrelevant content on Twitter. Lately, a new hashtag called “#whitelivesmatter” was used as a counter for another hashtag “#blacklivesmatter”. A lot of anti-government protests and various other violent activities were conducted, recorded, and posted on Twitter with this hashtag. A lot of Kpop fans had taken over this hashtag and flooded Twitter with extremely irrelevant content. Due to which the main and important content of the protests was drowned in these irrelevant tweets, which made it extremely hard for the officials to find and reinforce the law and order. This paper aims at building a model that helps in finding the relevance of text content in the tweet and its hashtag #whitelivesmatter in specific. In this paper, supervised data analysis techniques like text classification are used to get the required output.

Jyothirlatha Kodali, Vyshnavi Kandikatla, Princy Nagati, Veena Nerendla, M. Sreedevi

Smart Farming System Using IoT and Cloud

Due to the unprecedented increase in human population, agriculture plays an indispensable role in satisfying their daily needs. Henceforth, improving farm productiveness is indeed a huge challenge in the existing farming industry, which lacks continuous record management to satisfy the constantly emerging food needs. Along with the increasing population, global warming and climate transition also remain as an increasing challenge in the agricultural sector. In this scenario, this research has attempted to develop a smart farm management method, which incorporates cloud as well as the Internet of things (IoT) to take appropriate action. For instance, smart farming helps to provide a variety of important data such as air temperature. The paper has provided a smart device for farm field tracking, which controls dry run, motion detection, soil moisture detection, rainwater detection, humidity, and temperature. Also, this research work implements proper measures for those concepts on receiving the collected information without human input and later the detected quantities are stored for further data analysis within the cloud. Real-time feeds are being supervised upon this webpage as well as on the cell phone messaging. These would encourage farm workers and cultivate agricultural crops in a more modern way.

Neha Patil, Vaishali D. Khairnar

Multipartite Verifiable Secret Sharing Based on CRT

Subrahmanyam, Rolla Rukma Rekha, N. Subba Rao, Y. V.In (t, n) threshold secret sharing scheme, the dealer distributes secret among a group of n participants, and any t threshold number of participants can reconstruct the secret. However, $$t-1$$ t - 1 or lesser number of participants can not retrieve the secret. In verifiable secret sharing schemes (VSSS), participants can verify their share after receiving shares from the dealer to ensure that a dealer is not malicious. In multipartite secret sharing based on CRT scheme, a set of participants is divided into disjoint partitions, and whatever action performed at a single level is repeated at all other partitions. However, till date there is no mechanism to verify if dealer is malicious or not in multipartite secret sharing based on CRT. Two schemes are proposed for verification of a dealer, namely multipartite verifiable secret sharing based on CRT by using Iften, and multipartite verifiable secret sharing-based CRT by using kameer Kaya. Both proposed schemes are perfectly secure, and the security of both the schemes dependent on discrete logarithm problem.

Rolla Subrahmanyam, N. Rukma Rekha, Y. V. Subba Rao

Implementation of Smart Parking Application Using IoT and Machine Learning Algorithms

By considering the ever-increasing traffic in metropolitan areas, vehicle parking has become a great hindrance, especially while finding the available parking space nearby any office space or shopping mall, which is located on the narrow roadways. As the attempt to manually search for a parking slot consumes more time, commercial parking slots are designed to balance the demand and availability of vehicle parking spaces. Since constructing and monitoring a private parking space requires more money and workforce, parking charge has become very expensive. Due to the non-affordability of drivers, they waste more time in looking for empty parking slots. To overcome these challenges, the proposed research work helps to automatically identify the empty parking spaces, so that the car can be parked even in the most comfortable spot via video image processing and neural networks techniques, which develops a parking management software that actually identifies the existence of parking areas. The data from video footage is used to train the Mask R-CNN architecture, where a computer vision image recognition model is used to automatically identify the parking spaces. To label the car parking place mostly on the source images of a whole parking lot, a pre-processed region-based convolutional neural network (Mask R-CNN) is used. All of this could be solved by impelmenting a smart application, which could also send a text information to the customer, whenever a parking slot becomes available. Only at end of the day, it is required to have an appropriate and possible approach for solving all parking issues in and around the neighbourhood.

G. Manjula, G. Govinda Rajulu, R. Anand, J. T. Thirukrishna

Deep Face-Iris Recognition Using Robust Image Segmentation and Hyperparameter Tuning

Brown, DaneBiometrics are increasingly being used for tasks that involve sensitive or financial data. Hitherto, security on devices such as smartphones has not been a priority. Furthermore, users tend to ignore the security features in favour of more rapid access to the device. A bimodal system is proposed that enhances security by utilizing face and iris biometrics from a single image. The motivation behind this is the ability to acquire both biometrics simultaneously in one shot. The system’s biometric components: face, iris(es) and their fusion are evaluated. They are also compared to related studies. The best results were yielded by a proposed lightweight Convolutional Neural Network architecture, outperforming tuned VGG-16, Xception, SVM and the related works. The system shows advancements to ‘at-a-distance’ biometric recognition for limited and high computational capacity computing devices. All deep learning algorithms are provided with augmented data, included in the tuning process, enabling additional accuracy gains. Highlights include near-perfect fivefold cross-validation accuracy on the IITD-Iris dataset when performing identification. Verification tests were carried out on the challenging CASIA-Iris-Distance dataset and performed well on few training samples. The proposed system is practical for small or large amounts of training data and shows great promise for at-a-distance recognition and biometric fusion.

Dane Brown

Text-Based Sentiment Analysis with Classification Techniques—A State-of-Art Study

Social media acts as a bridge between people to widely share any data and communication. In recent years, the textual data content is increasing rapidly, where the text can contain any kind of information about people, product or service. Manually reading each text from online is not possible, and also, it is a challenging task to decide whether the user has positive stance or negative stance on the topic. To solve this problem, text processing techniques and algorithms are required. Sentiment analysis is the technology that processes any online text and classifies it into positive, negative and neutral. To analyze online content, new models are proposed by incorporating the machine learning concept. The unstructured information from online documents is analyzed and classified as results, which has been described as user sentiment analysis. The outcome of sentiment analysis can be used for business development, understand the customer expectations and also to know the public opinion toward a particular product or service. This paper focuses on various sentiment analysis processes and also the most used classification techniques from machine learning concepts.

M. S. Kalaivani, S. Jayalakshmi

Face Mask Detection Using MobileNetV2 and Implementation Using Different Face Detectors

Face recognition and object detection have been around the artificial intelligence field for a couple of years now and are constantly evolving and being pushed in many devices which we might not be even aware of. Many of such face and object detection techniques use Convolution Neural Network (CNN) architecture at the core for understanding and classifying any image passed on to the system. The neural networks identify many characteristics and distinguishing features present in the image and then provide us with a prediction. In this paper we discuss the implementation of a face mask detection technique using Mobile NetV2, observe the accuracy of our model and compare the performance of the trained model by incorporating three different face detector models. The result achieved from the trained model brings forth the opportunity for implementing such techniques on low computational powerful devices thereby making mask detection algorithm integration much easier than other techniques.

Kenneth Toppo, Neeraj Kumar, Preet Kumar, Lavi Tanwar

Image Encryption Using Diffusion and Confusion Properties of Chaotic Algorithm

Based on Chaos theory, cryptographic techniques show several new and successful ways to build reliable image encryption schemes. We present an image encryption in this paper using a logistic map of 1D. The proposed model framework is based on a main stream generator for the mechanism of uncertainty. A secret key of 256 bits, which is itself created by a logistic map, initiates the confusion process. In order to make the cipher more dynamic against any attack, after encrypting each block of the image, the secret key is changed. The experimental results show that the proposed approach offers an effective and safe way to encrypt and transfer images in real-time.

J. N. Swaminathan, S. Umamaheshwari, O. Vignesh, P. Raziya Sulthana, A. Hima Bindu, M. Prasanna, M. Sravani

A Sentiment Analysis of a Boycott Movement on Twitter

Sentiment analysis refers to determining emotional content from a textual input. The Internet world is now run by the term “Web 2.0,” and one of the major platforms that made this web to web 2.0 is Twitter, a microblogging social network where one can post short messages known as tweets. The general public uses Twitter social networking platform very often to post opinions on various topics ranging from reviews to current affairs. And one of the trending topics in India in the year 2020 was the boycotting online shopping services such as Amazon and Flipkart due to issues related to political clash between India and China and also issues related to nepotism in Bollywood. The border issue between India and China was a big talking point as it involved clashes between the Chinese and Indian soldiers which resulted in few martyrs. This made a major public anger toward China and its products which then resulted into the boycott movement. The other talking point was the nepotism in Bollywood. The recent suicide of Sushant Singh Rajput has brought out the public sadness and blamed Bollywood nepotism as the result of his suicide. This resulted in public boycotting products that were endorsed by actors/actresses that were having a favoritism which was granted by their relatives in the Bollywood industry. Hence, we decided that it was a great opportunity to extract tweets related to this boycott movement and do sentiment analysis on it to determine various emotions expressed by the general public.

Sooraj Bhooshan, R. Praveen Pai, R. Nandakumar

Implementing the Comparative Analysis of AES and DES Crypt Algorithm in Cloud Computing

Cloud computing helps users to access information over the internet. Efficiency, scalability and resource consumption are all optimized in cloud computing. Data security is an important concern in cloud computing that can be addressed with cryptography. Cryptography can be defined as a measure to protect confidential information that must be protected from others who are not intended to be viewed in the same way. The use of cryptography is widespread from small schools and colleges to the social media platform that we addicted to. AES and DES cryptographic algorithms are the main concern of this paper and also implementing which algorithm is best in the basis of time and also comparing the differences between these two algorithms.

R. S. Reshma, P. P. Anjusha, G. S. Anisha

A Model for Predictive and Prescriptive Analysis for Internet of Things Edge Devices with Artificial Intelligence

In recent years, abundant amounts of data have been accumulated from a huge network of Internet of things (IoT) devices spread around the globe. The collected data is only useful if it creates an action. To forge data actionable, it needs to be broadened with context and creativity. Traditional methods of evaluating structured data and creating action do not contribute to efficiently process the massive amounts of real-time data that stream from IoT devices. The study has shown that most of the IoT gadgets offering cloud storage along with analytics either trade the data or are lost dumped with no use. For instance, consider the trillions of log files that contain metadata, timestamps of a smart bulb which seems useless if used by nobody. But, it is always important to correlate the data with similar data patterns in a different application that helps in forecasting an insight into possible outcomes. Hence, there is a huge scope for improvement in this realm which motivated us to perform experiments and prove the concept with rigid conclusions. This is where AI-based analysis and response become crucial for extracting optimal value from that data. Also the research involved contains sensible prescriptive analysis offering hindsight when one talks about the edge or node devices in the IoT scenario but certainly, it lacks the rigid structure for offering insight and foresight. An in-depth insight at the edge level can be conceived by the existing artificial intelligence building models offered by many IT giants such as AWS Greengrass. Thus, there is an immense need to process the edge device data with enough intelligence and use existing analytics tools to greatly enhance the performance of the cloud and improve overall IoT application in hand by making the cloud requirements less CPU intensive and more economic. In this paper, a model for predictive and prescriptive analysis to improve production capabilities, gain efficiencies, and reduce operating costs by delving into edge computing to produce actionable insight and foresight is demonstrated with the help of a practical experiment.

Dinkar R. Patnaik Patnaikuni, S. N. Chamatagoudar

Comparative Analysis of SIM-Based Hybrid Modulation Schemes Over Log-Normal Channel Model

To enhance the performance of free space optical systems, the hybrid modulation scheme has been proposed. For this analysis, the performance of different hybrid schemes has been studied in the terms of the calculated bit error rates. In this paper, SIM modulation of hybrid schemes over log-normal model has been studied. Analysis of PPM-FSK-SIM, PPM-BPSK-SIM and PPM-GMSK-SIM-based hybrid modulation schemes over the log-normal channel model has been done using MATLAB software. The performance of PPM-GMSK-SIM is proved to be better than the performance of the other two schemes in the obtained results. Although other modulation schemes may have their own advantages, PPM-GMSK-SIM scheme proposed in this paper is specifically suited to free space optical systems. The variation of BER of PPM-GMSK-SIM with the various parameters such as atmospheric turbulence and link distance has also been analyzed in this paper. In addition, the detailed comparative analysis of different schemes presented in this paper can help in choosing the appropriate modulation technique in accordance with the desired application.

Siddhi Gangwar, Kavita, Subhash Burdak, Yashna Sharma

Lesion Preprocessing Techniques in Automatic Melanoma Detection System—A Comparative Study

An automatic melanoma detection system is an image processing-based technique used to detect melanoma. From the infected skin area image, the automatic melanoma detection system produces classification results as benign or melanoma. The automatic melanoma detection system contains four steps. Preprocessing step removes the noise from the infected image. The segmentation step finds the region of interest. Feature extraction is used to obtain lesion features and the classification step predicts lesion image as benign or melanoma. The decision of melanoma detection from such an automatic system depends upon the quality of the input image. Therefore, preprocessing of lesion images is an essential step in the automatic melanoma detection system. It becomes a challenging task due to the presence of various outliers like glare, dust, and hairs on skin lesions. Preprocessing techniques are applied for noise and artifact removal from the lesion. A lot of preprocessing techniques are available in the literature. The selection of appropriate preprocessing techniques may improve the accuracy of the automatic melanoma detection system. Therefore, in this work, we have studied and compared different preprocessing techniques so that the researchers may select appropriate techniques for them. This paper highlights and compares the image enhancement preprocessing techniques based on SNR and PSNR using pepper, salt, and Gaussian noise.

Shakti Kumar, Anuj Kumar

A Comparatıve Analysis on Three Consensus Algorithms

Proof of Burn, Proof of Elapsed Time, Proof of Authority

Blockchain technology has attracted immense attention in recent years from research, business, and governments all over the world. It is regarded as a mechanical advancement that is supposed to disrupt a few application areas that come into contact with all aspects of our lives. Nonetheless, a significant number of these blockchain implementations suffer from genuine flaws in their presentation and security, which must be addressed before any wide-scale acceptance can be achieved. The consensus calculation, which determines the exhibition and security of any blockchain system, is a key element. An algorithm for consensus is a tool that allows clients or machines to coordinate themselves in a decentralized environment. It must ensure that all nodes in the network will agree on a single source of truth even if a few nodes fail. They can usually make improvements, but there is not a complex administration structure in place to reach an agreement among multiple administrators. In a decentralized system, it is an entire other story. Assume we are dealing with a distributed database—how can we decide which records to add? Conquering this test in an atmosphere where nodes do not trust each other was perhaps the most demanding breakthrough in blockchain preparation. In this article, we will investigate how consensus calculations that are fundamental to the working of cryptographic forms of money and distributed ledgers. Therefore, a few current new agreement calculations have been presented to resolve the impediments of different blockchain structures. A methodical investigation of these equations can help us see how and why the way it operates is carried out by a particular blockchain. Consequently, in order to address the impediments of the various blockchain frameworks, some of the existing calculations of the new consensus have also been presented. A methodical investigation of these calculations will help to see how and why a specific blockchain plays out the way it works. In this article, we are talking about three consensus algorithms using an exhaustive empirical classification of properties and looking in depth at the implications of the various issues still prevalent in the agreement calculations.

Aswathi A. Menon, T. Saranya, Sheetal Sureshbabu, A. S. Mahesh

Issues and Challenges in the Implementation of 5G Technology

The next-generation mobile communication network (5G) is a heterogeneous network, and it has the added advantage in the wireless communication field. Users will feel uninterrupted communication over the 5G network. It required a higher bandwidth in order to achieve a higher data rate. 5G is classified into three categories such as ultra-reliable low latency communication (URLLC), massive machine-type communication (mMTC), and enhance mobile broadband (eMBB) by the International Telecommunication Union (ITU). 5G will provide the higher data rate (Gbps), low latency, enhance quality of service (QoS), low energy consumption at a low cost per transmission, better spectral efficiency (SE), energy efficiency (EE), quality of service (QoS), improved throughput and better user experience. There will be so many challenges to achieve the above-mentioned factors. The main challenges are to reduce Interference, latency, power consumption, and enhance data rate. The paper highlights the different issues and challenges of 5G and compare different existing methodologies for mitigating these challenges.

Mithila Bihari Sah, Abhay Bindle, Tarun Gulati

IoT-Based Autonomous Energy-Efficient WSN Platform for Home/Office Automation Using Raspberry Pi

In this fast-paced world, every person likes to work in a faster way to satisfy their needs. IoT helps to accomplish this goal by connecting large number of devices as an automated system so that the extra works of humans are reduced. Smart home is one the familiar example for IoT technology. The huge development in IoT technology and the support of existing automation techniques are utilized in this research work to develop a home automation system. Raspberry pi-3 along with google assistance is incorporated in the proposed design to control the electrical devices in the home from anywhere. Proposed model is suitable for elder persons who cannot able to switch on or off the appliances so that it can be controlled over voice. Also it reduces the power consumption efficiently.

M. Chandrakala, G. Dhanalakshmi, K. Rajesh

Detection of Early Depression Signals Using Social Media Sentiment Analysis on Big Data

Social media have become the new ‘reality’ for people as years go by and they have started linking their lives with these electronic devices. As a result, the increased chances of expressing themselves through media like Twitter, Instagram, Facebook, etc., have contributed to the study of depression analysis. The proposed paper predicts early signs of depression using supervised machine learning based on Naive Bayes, Decision tree, SVM, k-nearest neighbors on big data to find the accuracy on prediction.

Shruti S. Nair, Amritha Ashok, R. Divya Pai, A. G. Hari Narayanan

Raspberry Pi-Based Heart Attack and Alcohol Alert System Over Internet of Things for Secure Transportation

In this modern era, the number of accidents that occur around us is increasing unprecedently. The reasons for accidents might be the driver’s health issues, drowsiness and alcohol consumption. To prevent accidents and improve the driver’s safety, this research work has proposed a smart and secure transportation system. The proposed paper describes the design and implementation of smart and secure transportation system by using Raspberry Pi and Internet of Things. For alcohol detection, MQ3 sensor is used, and to detect heart attack, pulse sensor is used. Here, a gadget has been proposed with the intention to discover coronary heart assault via tracking the heart rate based on Internet of Things (IoT). Monitoring and alerting an individual, when the comatose gains cognizance by using the movement detection gadget. Two sensors are used to monitor the health of the patient. Pulse sensor is used to monitor the pulse rate functioning of heart. Alcohol sensor is used to detect alcohol percentage in air. By using these sensors, the physical condition of the driver can be detected. The system locks the vehicle if the driver is determined to be alcoholic or with improper pulse rate. Hence, it prevents the road accidents.

G. Dhanalakshmi, K. Jeevana Jyothi, B. Naveena

Automatic Classification of Music Genre Using SVM

The growing number of music content online has opened up new possibilities for the introduction of successful digital knowledge access services known as music referral systems that help user groups in searching, finding, sharing, and creating. The music recovery approach based on specific similarity information combines several similarity features, including audio and contextual similarities, such as tone format features and melodic details. Audio classification is very important for recovering audio files quickly. To get the best results from audio classification, it is important to choose the best feature set and follow the best analysis method. Support vector machines (SVMs) are implemented by learning from input samples to classify music into separate classes of music genres. The SVM study excelled in the music category classification.

Nandkishor Narkhede, Sumit Mathur, Anand Bhaskar

Crime Rate Prediction Based on K-means Clustering and Decision Tree Algorithm

The major cause of crimes that initiate nuisance for society in many ways is human behavior disorder. In many countries, the crimes and accidents are seriously monitored. Crime analysis is a process, which completely analyses the patterns and trends over a period of time. Nowadays, different sources of crime data provide a greater opportunity for performing large analysis in the research community. The proposed work analyses the crime under different locations (considering latitude and longitude) and different time periods. The proposed work predicts the crime type and gives an input about the date and location. The application is developed as a Windows application by using TKinter-Python for crime prediction. Machine learning concepts and implementation are used here for performing crime analysis and prediction, which aid the ease of understanding the data in multiple ways and further predict it with good accuracy. The algorithms used behind this work are K-means and decision tree algorithm. The proposed work aims at analyzing the data mining concepts for clustering and classifying the crime prediction. The results show that the classification method outperforms in terms of detection and accuracy. Experimented results are evaluated for error calculation, and they are further analyzed in this study.

Jogendra Kumar, M. Sravani, Muvva Akhil, Pallapothu Sureshkumar, Valiveti Yasaswi

Comparative Study of Optimization Algorithm in Deep CNN-Based Model for Sign Language Recognition

The fundamental part of the neural network is the learning rate, and the strategy of adopting the learning process in a neural network is carried out using optimization algorithms or optimizers. This optimization algorithm helps us produce better results to the model by changing the parameters like bias and weights, i.e., it helps us maximize or minimize the error function and depends on the learnable parameters. In this paper, we examine how an End-to-End CNN model named ASLNET recognizes the alphabets of the American sign language using various optimizers such as Stochastic Gradient Descent (SGD), Root-Mean-Square propagation (RM-Sprop), Adaptive Gradient Algorithm (Adagrad), Adaptive Delta (Adadelta),Adaptive Moment Estimation (Adam), Adam with Nesterov Momentum (Nadam), LookAhead and Rectified Adam (RAdam). To avoid the overfitting issues, traditional data augmentation techniques are used to compare our model with data augmentation and without augmentation with these optimizers. Among these, LookAhead and RAdam are the most recently developed. The experiment is conducted on 2 NVIDIA TESLA P100 GPUs of batch size 64, and the investigation was based on benchmark ASL Finger Spelling dataset.

Rajesh George Rajan, P. Selvi Rajendran

Cardinal Correlated Oversampling for Detection of Malicious Web Links Using Machine Learning

The problem with malicious websites is growing day by day as it leads to the black listing of websites. The unauthorized websites are gathering the user’s database information and their assets. Few of the URLs are completely used as a host webpage to publish unrelated web content that signifies cyber-attacks. Cracking the presence of malicious website still pertains as open task due to the lack of web characteristics for malicious and benign websites. To overcome this problem, we are using machine learning techniques for detecting the malicious content and web links. Backgrounding the above, this paper used malicious webpage dataset extracted from UCI dataset repository for predicting the level of mushroom edibility. The categorization of malicious webpage classes is achieved in five ways. Firstly, the dataset consisting of 21 features with 1781 records and is preprocessed with encoding, feature scaling and missing values. Secondly, raw dataset is fitted to all the classifiers with and without the presence of feature scaling and the performance is analyzed. Thirdly, the cardinality free malicious dataset is fitted to all the classifiers with and without the presence of feature scaling and the performance is analyzed. Fourth, the correlated free malicious dataset is fitted to all the classifiers with and without the presence of feature scaling and the performance is analyzed. Fifth, the oversampled malicious dataset is fitted to all the classifiers with and without the presence of feature scaling and the performance is analyzed with precision, recall, accuracy, running time and F-score. Implementation analysis portrays that the decision tree classifier for raw, Cardinality reduced dataset tends to retain the accuracy with 97.1% before and after feature scaling. The random forest classifier with correlated free dataset tends to retain the 96.6% accuracy before and after feature scaling. Decision tree classifier for oversampled dataset tends to retain the accuracy with 98.4% before and after feature scaling. From the above analysis decision tree classifier is found to be more efficient in its accuracy with all raw, cardinality free and oversampled dataset.

M. Shyamala Devi, Uttam Gupta, Khomchand Sahu, Ranjan Jyoti Das, Santhosh Veeraraghavan Ramesh

Simulation of Speckle Noise Using Image Processing Techniques

The image noise is considered as one of the significant problems in scientific applications. The simulation of the speckle noise within a standard image is studied using the presented algorithm. Different speckle noise ratios were added, with per cent (0.01–0.06), to simulate noise within different images. This added noise based on the mathematical equations to simulate the behavior of this type of noise. The main work divided into two steps; the first step is the classification method which based on the minimum distance and it used to classify the tested image with different homogenous areas and compare it with the corresponding noise images in the same location. In the first step, the knowledge of understanding the behavior of speckle noise achieved. The second step is the statistical criteria namely mean and standard deviation which is used to calculate the speckle factor (SF) to know the effect of noise within the image. In this step, the effect of the noise is obvious by checking the statistical values of SF. The behavior of the speckle noise is well-described and recognize based on the presented algorithm and method.

Noor H. Rasham, Heba Kh. Abbas, Asmaa A. Abdul Razaq, Haidar J. Mohamad

Wi-Fi-Based Indoor Patient Location Identifier for COVID-19

Nowadays, more number of people are affected by COVID by other infected people. The coronavirus spreads through an infected person when he/she talks, coughs or sneezes in front of others. Before getting COVID test results, infected person may be moving along with normal people in shopping malls, education campus and industries. It is difficult to identify uninfected people who are moving along with the infected people. This paper proposed location identification model which is used to track and locate the infected person or recovered person inside the building. The location sensing model in wireless network and global positioning system (GPS) cannot be used to track users inside the buildings. The proposed system uses Wi-Fi-based model for monitoring patients and identifying the location of infected people in different floors of the building. RSSI technology is used for tracking the mobile device that is carried by patients in the indoor environments. Then, it finds the other uninfected people near them by using their smartphones.

A. Noble Mary Juliet, N. Suba Rani, S. R. Dheepiga, R. Sam Rishi

Enabling Identity-Based Data Security with Cloud

In disseminated stockpiling organizations, customers store data indirectly onto the cloud and comprehend the data giving to others. Far away information uprightness taking a gander at is proposed to the attestation the respectability of the informational index aside in the cloud. In some fundamental appropriated amassing frameworks, for example, the EHRs structure, the cloud document may give touchy data. At the point when the cloud report is shared, the delicate information ought not to be known to anybody. Bringing the whole shared report can comprehend the inclusion of the collaborating records; however, it does not permit different gatherings to utilize this shared record instructions to recognize data offering to sensitive information stowing away in distant data uprightness exploring still has not been researched up to now. To address this issue, this paper proposes a distant data genuineness surveying plan that recognizes data offering to sensitive information concealing. To accomplish high security, this paper proposes security calculations for producing the private key and for making record label name for relating information. The entire information's are encoded twice prior to shipping off cloud. Just the information proprietor can see the entire information from cloud. Personality-based trustworthiness examining ensure that different clients could just see the required information by concealing delicate data safely.

Arya Sundaresan, Meghna Vinod, Sreelekshmi M. Nair, V. R. Rajalakshmi

A Study and Review on Image Steganography

Steganography is the science that involves encrypting data in a suitable multimedia carrier, such as image, audio, and video files. The main purpose of image steganography is to hide the data in images. This means that it encrypts the text in the form of an icon. Steganography is done when there is communication takes place between sender and receiver. In a day of data transfer over the network, security is paramount. Before the development of stenography, data security is a major research concern for researchers. Steganography is gaining importance due to the rapid development of users on the Internet and secret communication. In this paper, we discuss about various type of existing image steganography techniques and analyze the advantages and disadvantages of different types of image steganography techniques.

Trishna Paul, Sanchita Ghosh, Anandaprova Majumder

Fault Detection in SPS Using Image Encoding and Deep Learning

Satellite power system (SPS) is considered as the core of the satellite, where the faults occurring here adversely have an impact on the health of the satellite, thereby affecting the mission. This can be avoided by early detection of the faults occurring in the SPS. This work proposes a model to classify the faults present in the SPS using 2-dimensional convolutional neural network (2-D CNN) by encoding the multivariate time series data present in the ADAPT dataset into images. Encoding is done by using the methods such as Markov transition field (MTF), Gramian angular summation field (GASF), recurrence plot (RP), and spectrogram. Promising results were obtained using the GASF and 2-D CNN combination, which have yielded a test accuracy of 87.5%. The precision, recall, F1 score, and AUC score were 0.89, 0.854, 0.865, and 0.94, respectively.

P. Hari Prasad, N. S. Jai Aakash, T. Avinash, S. Aravind, M. Ganesan, R. Lavanya

A Comparative Study of Information Retrieval Models for Short Document Summaries

The judicial system has evolved tremendously over the past years. Thousands of cases are registered daily and stored in the form on documents which are used by lawyers whenever required. Lawyers are important stakeholders in judicial system and constantly study multiple cases during their work. Manual retrieval of this information from a collection is very difficult. This is where the information retrieval system comes in picture. This article is a brief comparison of various information retrieval models which are currently being used. It includes the Boolean model, TF-IDF model, vector space model, Okapi BM25 model and fuzzy search models. Each of these models is tested on three datasets, and their results were noted. The experimental results unfold that the Okapi BM25 model outperformed the other models in the case study. The results also show that document pre-processing plays an important role in the effectiveness of the query-document matching.

Digvijay Desai, Aniruddha Ghadge, Roshan Wazare, Jayshree Bagade

Network Attack Detection with QNNBADT in Minimal Response Times Using Minimized Features

Internet is a medium of globally interconnected independent networks. Though it was created to interconnect government research laboratories in 1994, it has witnessed phenomenal growth and has expanded to service millions of users in governments, academia and public/private organizations for multitude of purposes. Internet has been evolving continuously. Internet has also evidenced many attacks on its networks called cyberattacks. As Internet evolves, adversaries also evolve in their attacking techniques making it imperative to guard networks from attacks. In spite of firewalls, AVs (Anti Viruses) and other defense mechanisms, there is an implicit need to monitor deliberate proliferations. IDSs (Intrusion Detection Systems) are techniques that help monitor networks and raise alarms on finding damaging proliferations. This also implies IDSs need to be quick in their assessments of malicious behavior on the network. This paper proposes a NN (Neural Network) based IDS that can quickly respond to attacks by analyzing low-level network details. The proposed scheme is evaluated on the In CIRA-CIC-DoHBrw-2020 dataset where it averagely scores above 90% in accuracy when benchmarked on different sample sizes.

S. Ramakrishnan, A. Senthil Rajan

Deep Learning-Based Approach for Satellite Image Reconstruction Using Handcrafted Prior

We propose a randomly initialized neural network as handcrafted prior to distorted satellite image for its restoration. The model is applied for cloud removal and proved efficient. Extensive experiments on the satellite datasets demonstrate efficiency of the proposed model both quantitative and qualitative. Further, the proposed approach also removed the dependency on pre-training datasets. In our study, RGB monochromatic satellite images were considered with the obscured area of varying shapes, lying in the range of 14–30%. Reconstructed image with MSE 0.131 and PSNR of 80.937 is obtained. Another inference deduced from the results is structural symmetry index (SSIM) values are better for red and green bands when compared to blue band. Image hash value is also calculated and found satisfactory.

Jaya Saxena, Anubha Jain, Pisipati Radha Krishna

CLOP Ransomware Analysis Using Machine Learning Approach

Machine learning seems to be evolving day by day in various technological aspects. It has become a powerful tool that may use to do both harm and good. Data breaches, ransomware threats, Internet of Things (IoT), etc., are some of the threats faced by cybersecurity. Traditional cybersecurity methods could not tackle these attacks. Among malware, ransomware is a particularly diabolical type of malware. Once ransomware gets on your computer usually through an infected email attachment or all-too-common Trojan horse attack it will lock your computer or your data in some way and demand payment in exchange for giving control of your system back to you. Some simple ransomware model will simply try to fool the users and make them to spend more money for fixing it. Clop ransomware is considered as one of the most dangerous malwares. Most of the computers will become victims to Clop ransomware. Nowadays, it is an increasing concern among large companies. So, the main purpose of this study is to provide extreme surveillance, for that a survey has been proposed on Clop detection using machine learning methods. Using machine learning algorithms, the study analyzes clop detection in three different datasets using different machine learning algorithms and measured these conclusions with a similar malware detection study. Based on our evaluation, it is observed that XGBoost outperforms all other machine learning algorithms. The proposed study answers the question regarding the best machine learning algorithm for Clop detection.

E. S. Aiswarya, Adheena Maria Benny, Leena Vishnu Namboothiri

Integration of Wireless Sensors to Detect Fluid Leaks in Industries

The industrial internet of things is a specific domain that deals with industrial machines and their communications. This allows us to bring better reliability and efficiency in the work operations. To increase the reliability in the system, suitable error detection methods should be embedded in the system process. Design implementation and testing for the pipe leak detection are done in this paper. In this work, a prototype is created with small wireless nodes distributed along the pipelines, pressure sensors are installed next to the manually created leak nodes. The leak is detected using the pressure point analysis method. Various pressure data are obtained from the fluid flow networks in the pipes, and its variations are analyzed using Bernoulli’s equation, to monitor leaks in pipes.

N. Santhosh, V. A. Vishanth, Y. Palaniappan, V. Rohith, M. Ganesan

Performance Analysis of Abstract-Based Classification of Medical Journals Using Machine Learning Techniques

Researchers face many challenges in finding the opt web-based resources by giving the queries based on keyword search. Due to advent of Internet, there are huge biological literatures that are deposited in the medical database repository in recent years. Nowadays, as many web-based medical researchers evolved in the field of medicine, there is need for an intelligent and efficient extraction technique required to filter appropriate and opt literature from the growing body of biomedical literature repository. In this research work, new combination of model is proposed in order to find the new insights in applying the combination of algorithm on biological data set. The information in the biomedical field is the basic information for healthy living. National Center for Biotechnology Information (NCBI)’s PubMed is the major source of peer-reviewed biomedical documents for researchers and health practitioners in the field of health-related management. In this paper, abstracts available in PubMed database is used for experimentation. In recent years, deep learning-based neural approach models provide an efficient way to create an end-to-end model that can accurately measure classification labels. This research work is a systematic analysis of performance of the supervised learning models such as Naïve Bayes (NB), support vector machine (SVM) and long short-term memory (LSTM) by implementing on textual medical data. The novelty in this work is the process of incorporating certain topic modelling techniques after the pre-processing phase to automatically label the documents. Topic modelling is a useful technique in increasing the efficiency and improves the ability of researchers to interpret biological information. So, the classification algorithms thus proposed are implemented in combination with popular topic modelling algorithms such as latent Dirichlet algorithm (LDA) and non-negative matrix factorization (NMF). The final performance of the combination of algorithms is also analysed and is found that SVM with NMF outperforms the other models.

A. Deepika, N. Radha

Development of Improved SoC PTS Algorithm for PAPR Reduction in OFDM Underwater Communication

Orthogonal frequency division multiplexing (OFDM) is a multi-carrier communication technique used in many modern day communication platforms. It was designed with a very high data rate during wireless communication. Underwater communication is an emerging technology in the field of wireless communication. Due to the detrimental effect of time and frequency spreading, achieving good data transfer in underwater wireless communication is challenging. This paper proposes a multi-level optimization model called partial transmission sequence (PTS). This work utilizes PTS algorithm to reduce the peak-to-average power ratio (PAPR) of OFDM systems in underwater communication. This method increases the number of optimization stages and reduces the number of elements in the present phase set for each level of optimization to find the optimal phase rotation. The computational complexity of traditional PTS algorithm increases exponentially with the increase of the number of sub-blocks and the number of phase set elements can be selected. The proposed implementation would reduce the complexity of the system as the sub-blocks are executed in parallel. The development of proposed method is divided into two stages. The proposed PTS model is first developed in MATLAB and then implemented on system on chip (SoC) platform. The proposed method is compared with the existing methods. The experimental results prove that the proposed method greatly reduces the PAPR value, thereby increasing the overall performance.

M. Asha, T. P. Surekha

Analysis of Twitter Data for Identifying Trending Domains in Blockchain Technology

Opinion data collection is one of the most important forms of data analysis to understand and gain more insight about the trending information related to any domain or technology. The need for opinion data mining on the Twitter data is the demand of the indeed titled historical big data era. Blockchain is the technology which was introduced for cryptocurrency and later claimed to be embraced in most of the technologies because of its efficiency in ensuring privacy, security, and data management. With the increase in the popularity of blockchain technology, the opinion data collection for the blockchain technology is becoming compulsive to identify its substance in the practical application of different sectors. The utmost intention of this research is twitter data analysis centred on the domains that are believed to be domains that apply blockchain technology and hence ascertain that they are active and trending domains. The data analysis is performed on the tweets downloaded using tweepy API. This research engages different data visualizations, and Domain Identification by extracting features from tweets. The trend analysis for the opinion mining is accomplished by considering the re-tweets of the considered tweets. The proposed analysis is carried out on tweets which are carefully streamed using filter words related to the domain which claim to be active applicants of the blockchain technology. We will focus on techniques for the extraction of the Twitter data related to blockchain, processing, segregation, pattern visualization and trend identification by considering natural language processing paradigm.

Sahithya Mareddy, Deepa Gupta

Enhancing the Security for Smart Card-Based Embedded Systems

Nowadays, security has become an uncertain phenomenon in the modern world. User data can be stolen and other privacy risks are posing a major threat to the society. With the increase in design complexity and cost of setting up a foundry, it has led to globalization of integrated circuit supply chain which poses many security threats like piracy and hardware Trojans, which leads to data leakage to the outside world. In this paper, we are proposing a security method for a smart card-based embedded system. Remote user authentication and key agreement scheme for smart cards are the way to go. It is a very practical solution to validate the eligibility of a remote user and provide secure operation of the system, hence securing the user data from untrusted sources.

G. Kalyana Abenanth, K. Harish, V. Sachin, A. Rushyendra, N. Mohankumar

Implementation Mobile App for Foreign Language Acquisition Based on Structural Visual Method

Among all modern communication devices, mobile phones are the most powerful means of communication even richer than e-mail or chat, as it can act as a teaching device despite its technical limitations. Therefore, we need to create a fast and efficient automated system that allow increased interest of students to learn foreign language, which provides the cognitive activity of students, stimulates and develops cognitive processes: thinking, perception, and memory. With such a system, the learner controls the learning process and progress in his own space based on his cognitive state, and learners can speak the target language without effort and psychological barriers. The purpose of the research is to create interactive speech trainer system based on a structural and visual approach and to ensure the formation of stable foreign language skills of trainees on the background of the active use of visual representation of language and interactive speech technology, and this system uses a technique for applying the visual approach and structural visual method in the educational environment by transforming grammatical information from verbal to graphic form and replacing complex text rules with appropriate visual structures in the form of pictures, schemes, and diagrams. We present our steps to implement our proposed architecture based on a visual model as a platform in mobile application with the establishment of content management system to provide the process of controlling the formation of speech skills and allowing the transition from foreign language learning to its improvement and acceleration. We also describe the ideas that will guide the design of this system.

Imad Tahini, Alex Dadykin

Assessing Deep Neural Network and Shallow for Network Intrusion Detection Systems in Cyber Security

Intrusion detection system [IDS] has become a central layer that unites everything inside the most recent ICT structure on account of the consideration for advanced prosperity inside the ordinary world. Motivations to recall the weakness to search out the sorts of assaults and grow the intricacy of bleeding edge computerized assaults; IDS requires the need to hitch deep neural networks (DNN). During this report, DNNs will not foresee assaults on the N-IDS. A DNN with a learning pace of 0.1 is applied and runs for the assortment of 1000 years, and subsequently, the informational index KDDCup-‘99’ was utilized for readiness and site meaning association. For assessment purpose, the arrangement is finished on the comparable dataset with another obsolete AI figuring and DNN of levels begin from 1 to 5. The outcomes were broke down, and it had been accepted that a DNN of three levels would be advised for execution.

Deena Babu Mandru, M. Aruna Safali, N. Raghavendra Sai, G. Sai Chaitanya Kumar

Leveraging Association Rules in Feature Selection to Classify Text

Appropriate featureAghbari, Zaher Al selection is an importantSaeed, Mozamel M. aspect in the fields of data mining and machine learning. Feature selection reduces data dimensionality and produces simpler classification models that have lower variance. In this paper, we propose a robust feature selection method that produces smaller and yet more effective set of features. The proposed feature selection method leverages association rules to select the effective features for text classification. Our experiment shows that the proposed method outperforms its pears in terms of execution time and classification accuracy.

Zaher Al Aghbari, Mozamel M. Saeed

Protected Admittance E-Health Record System Using Blockchain Technology

E-health record is considered as an individual’s health document, which gets shared among several amenities. The EHR system is becoming a popular protagonist as it has the potential to transform the paper-based industry into the digital system for maintaining the patient’s health records. Nevertheless, the E-health record system needs renovation in terms of confidentiality and access to records since the hospital authority is treated as a central holder for the records. The proposed system comes up with a novel solution to modernize the traditional centralized system by the erection of a decentralized framework, whereas the patient is the sole owner of the health document. Implementation of blockchain technology with a decentralized patient-centered structure promotes a secure healthcare system. Blockchain’s open admittance would permit changes to an individual’s her, which needs to be restructured in real-time and makes it instantly available to parties involved for the same. The use of shrewd agreements and dispersed stockpiling improves the therapeutic administrations, including clinical records comparably persistent related proof. This high-level model gives high security and ease of utilizing the highlights through disseminated record where information can’t be held for payment, where every client has a refreshed duplicate of the blockchain.

Sharyu Kadam, Dilip Motwani

Using Hierarchical Transformers for Document Classification in Tamil Language

Riyaz Ahmed, M. Raghuraman, Bhuvan Briskilal, J.Document classification is used for various applications from spam detection in email to article classification. Recently document classification in Tamil has been gaining momentum due to increased data available in said language. One of the major advances in document classification is due to bidirectional encoder representations from transformers (also known as BERT), which uses transformer architecture and has been used effectively in many natural language processing problems like sentiment analysis and document classification for Tamil language. One of the main disadvantages of pre-trained BERT model is the number of tokens cannot be higher than 512; otherwise, it has to be retrained. Our implementation mitigates this issue by using hierarchical transformer architecture and is especially useful for resource poor languages like Tamil. We compare hierarchical transformer model and compared with classical machine learning algorithms and found recurrence over BERT shows substantial improvement over SVM, logistic regression and random forest, with a weighted average F1 score of 0.88 for news article classification.

M. Riyaz Ahmed, Bhuvan Raghuraman, J. Briskilal

Analysis of Hybrid MAC Protocols in Vehicular Ad Hoc Networks (VANET) for QoS Sensitive IoT Applications

Vehicular Ad Hoc Networks (VANET) represent the base for future Intelligent Transportation system (ITS). The technology proposed to face different traffic problems related either to human safety, general safety, or Internet-related services. MAC protocols and routing protocols were developed to play a vital role in VANET communication. MAC protocols represent the time organizer to utilize the channel efficiently. This paper introduces the concept of TDMA efficiency working under hybrid protocols. It discusses recent hybrid MAC schemes and their performance in different proposed scenarios as well as in the term of TDMA frame efficiency. Efficiency of the frame serves in the enhancement and incrementing the bits in one TDMA frame. Where the efficiency is the amount of useful data to be transmitted within an allocated time can be expressed as the efficiency of access method utilized in MAC layer. Efficiency of TDMA frame represents the ratio of useful information bits to the total bits in a frame. This paper discusses hybrid MAC protocols with respect to QoS parameters. It also provides a comparison between the efficiency of hybrid MAC protocols.

Nadine Hasan, Arun Kumar Ray, Ayaskanta Mishra

Programming with Natural Languages: A Survey

Thomas, Julien Joseph Suresh, Vishnu Anas, Muhammed Sajeev, Sayu Sunil, K. S.Programming with natural language is a research area that has a wide range of applications including basic programming, robotics, etc. Factors like preserving the meanings, handling the ambiguity, etc. have to be considered while converting a natural language text to programming language statements. Many developments have been taken place in this area over the past few years. Different types of CFG parsers were used initially for converting the natural language texts to programming language statements. The developments in technologies based on AI have a huge impact in this area and efficient models like GPT-3 are created. These models are capable of converting natural language to target programming language more precisely. In this paper, we do a detailed and systematic study of the developments that happened in this area and list some of the relevant research works among them.

Julien Joseph Thomas, Vishnu Suresh, Muhammed Anas, Sayu Sajeev, K. S. Sunil

An Exhaustive Exploration of Electromagnetism for Underwater Wireless Sensor Networks

The influence of underwater sensor network (UWSN) has major role in the development of many areas like oil and gas, pollution, surveillance and marine life. The major success of UWSN is its ability in real-time monitoring, measurement, analysis, storage and transmission of the collected data to the outside world. The real-time monitoring of the measured parameters and the ability of transmitting huge amount of data from underwater to the terrestrial wireless sensor network is possible with the backup of technology development like Internet of submerged things (IoST). One of the factors that determines the quality of network is being reliability, and the deployment of sensors in dynamic condition is the major challenge in UWSN. Through this paper authors perform a thorough examination of the possibility of RF communication through the dynamic RF UWSN.

Nebu Pulickal, C. D. Suriyakala

Two-Stage Feature Selection Pipeline for Text Classification

Text classification is one of the significant fields in NLP. It has numerous applications in the business world such as e-mail spam filtering, fraud detection, recruitment and various other tasks. The past three decades have witnessed an exponential rise in the amount of information. This has made text classification all the more challenging. There is a greater need for optimal feature subset selection which in turn enhances classification performance. In this study, we discuss a two-stage feature selection pipeline which combines conventional filter methods (Chi square and information gain) with evolutionary algorithms (particle swarm optimization and genetic algorithm). Subsequently, we aim to compare the results from these pipelines with the results from evolutionary algorithms over existing classifiers. The experiments are performed on two popular data sets—20 newsgroups and IMDB reviews. Results obtained show that the pipeline-based feature selection methods perform better than their respective wrapper methods.

Vinod Kumar, Abhishek Sharma, Anil Bansal, Jagnur Singh Sandhu

A Systematic Approach of Analysing Network Traffic Using Packet Sniffing with Scapy Framework

Many professionals and network engineers face many issues where the available resources have been very unsuccessful in solving the problem. Tools which are available will provide better outcomes; however, it generates large amount of information chunks that requires time and effort to remove those unwanted data. Therefore, learning some scripting often helps to solve problems, analysis and carry on with automation helps to save time, expense and effort. This research work discusses the development in packet sniffer through Python along with Scapy framework. Packet sniffer is a network traffic and data interception, tracking, and analysis software tool. Including a sniffer research work shed light on packet elements, packet layers, designing elements, dissection and sniffing along with further exploration.

S. H. Brahmanand, N. Dayanand Lal, D. S. Sahana, G. S. Nijguna, Parikshith Nayak

Detecting Ransomware Attacks Distribution Through Phishing URLs Using Machine Learning

During the last few years, phishing and ransomware seem to be the most widespread type of threat and a rapidly growing hazard to organizations. Nowadays, ransomware is injected via phishing emails, poorly informed users appear without much thought to click on links and attachments, believing that the emails are valid hence any defence action must concentrate on e-mail security. Ransomware is a malicious code or software that encrypts files in the victim’s system in a simplified way, and the perpetrators are demanding a lot of ransom for it. Ransomware can inflict enormous harm to companies, resulting in a lack of production and sometimes economic difficulties. Most importantly, there is a lack of documents that may reflect hundreds of hours of work or customer information that is important to the effective operation of the organization. Traditional defence methods that depend on malware signatures and fundamental protection rules have been found to be ineffective towards ransomware threats. In reality, attackers are developing their malware to compromise conventional Web and email security set-ups that are vulnerable. A thorough review of the organization’s defensive measures should address the ransomware problem and learn how they are truly capable of reacting to the current threats. The paper goal is to see how using this approach helps to prevent malicious malware and uses it as a self-defence tool through machine learning techniques. Research was done on several uniform resource locator (URL) samples and the findings show that we can say the difference between malicious and benign URL.

B. N. Chaithanya, S. H. Brahmananda

A Framework for APT Detection Based on Host Destination and Packet—Analysis

In cybersecurity, advanced persistent threats have gained more attention. Even after that, a variety of techniques, like change control, sandboxing, and internet traffic analysis, are often used to identify APT attacks. Even so, 100% accomplishment wasn’t achievable. APTs use a variety of sophisticated methods to overcome several types of detection, as per recent research. This paper examines the most standard strategies, techniques, and mechanisms used by adversaries to describe and evaluate APT problems. It also outlines the vulnerabilities and capabilities of current security technologies that are in usage from the time the threat was detected in 2006 till now. Besides, this study introduces a different mechanism to eliminate the issue through APT with internet traffic by host destination and packet inspection.

R. C. Veena, S. H. Brahmananda

A Trust-Based Handover Authentication in an SDN 5G Heterogeneous Network

The fifth generation (5G) of wireless network paves the way for the development of new technologies to overcome the existing challenges in a heterogeneous network (HetNet). 5G supports huge data traffic with fastest and reliable network access. The main challenge in the existing 5G HetNet is the presence of different types of cells installed in the same geographical area. The frequent handover and authentication among different small cells gives rise to security challenges like access point insecurity, handover vulnerability and attacks. To overcome the existing challenges in such vulnerable network, software-defined networking (SDN) is introduced which is found to reduce the complexity of 5G networks and construction cost. Hence, this paper proposes a SDN-based handover authentication to enable efficient handover authentication and to enhance the security in a 5G mobile communication. The proposed algorithm helps to achieve mutual authentication using a three-way handshaking protocol. Moreover, the security of the proposed authentication scheme is evaluated using a trust value algorithm with clustering mechanism. From the experimental results, it is found that the performance of the proposed mechanism is better in terms of throughput, packet delivery ratio, and reduced delay when compared to the existing system.

D. Sangeetha, S. Selvi, A. Keerthana

Intrusion Detection for Vehicular Ad Hoc Network Based on Deep Belief Network

There has been continued to be a lot of research into self-driving and semi-self-driving in the last few years, which has to the creation of the vehicular ad hoc networks, but has become more vulnerable to potential attacks due to the misuse of networks. The proposed model of deep learning algorithm, namely deep belief network is used for detecting intrusion in the vehicular ad hoc network (VANET). Deep belief network algorithm gives more accuracy for intrusion detection in the network than existing methodologies such as machine learning algorithms or another deep learning algorithm. Nowadays, automation is more important in all fields, similarly automatic vehicles, i.e., driverless cars. These types of vehicles will come to market and all these vehicles are connected through a wireless network. All the vehicles are communicating with each other by sending some informative packets but there is an attacker who accesses that data and changes the data which may affect the security of the vehicle and also damage the system responsible for the accident. So, intrusion detection system for the vehicular ad hoc network is important with maximum accuracy. For this purpose, we used the updated CICIDS2017 dataset for training, testing and evaluation process. Experimental results using a deep belief network for intrusion detection mechanisms proved that the proposed model could have good results on multiclass and binary classification accuracy 90% and 98% respectively.

Rasika S. Vitalkar, Samrat S. Thorat, Dinesh V. Rojatkar

Highly Secured Steganography Method for Image Communication using Random Byte Hiding and Confused & Diffused Encryption

This paper propounds a new notion for a secured method for image communication using random byte hiding (RBH) technique with confused and diffused data embedding technique. This entire method is based on dual keys for embedding as well as for retrieving. Security through obscurity followed by Kerckhoff’s principle is the main ideology of this method. Because of two keys, a large keyspace is needed, which can be effigiated only by brute-force attack. The cover image can be retrieved by using one key, and the secret image can be retrieved by using another key. Among the other dual-key encryption techniques, this method upholds its advantage in the form of security since the dual-key concept is used for two encryptions, which are very difficult to predict and break. The main motivation of the proposed technique is to reduce the time consumption along with increasing the security. The RBH algorithm reduces the encryption time and decryption time by 87.9% and 67.04%, respectively, compared to the conventional LSB steganography. The data hiding rate can also be improved to an extent of 98.33% compared to the conventional LSB technique.

S. Aswath, R. S. Valarmathi, C. H. Mohan Sai Kumar, M. Pandiyarajan

An Enhanced Energy Constraint Secure Routing Protocol for Clustered Randomly Distributed MANETs Using ADAM’s Algorithm

Mobile ad hoc network (MANET) contains association attributed to autonomous enforced dynamically changing mobile nodes that forms a dynamic network beyond having fixed network infrastructure. In the mobile ad hoc network, each and every mobile node will be operated with battery sources and each mobile node consists of limited energy. Various routing protocols are developed for routing, but every scheme suffers from the power constraint. The proposed scheme gives a solution for manipulating diverse paths among destination and source to reduce power constraints. The major intention is for finding best route among available routes between source and intermediate target mobile nodes. The energy model is considered to enhance the energy in routing protocol. Here, the ADAM’s algorithm is devised for implementing power aware routing in the MANET environment. At the beginning, the dynamic mobile nodes are compiled and simulated in a randomly dispersed domain for persuading energy constrained secure routing, where the power in each and every mobile node is being computed. Once the energy levels of all the mobile nodes are evaluated, then the secure routing scheme is introduced and settled by using dynamic secure nodes. Thus, the secure routing scheme is developed by using the recommended ADAM’s model. Thus, by considering the ADAM’s algorithm, the most favorable way of information routing is continued in MANET ambiance.

Bandani Anil Kumar, Makam Venkata Subamanyam, Kodati Satya Prasad


Weitere Informationen