Cybersecurity and Evolutionary Data Engineering

Today, data and its security are utmost important for all organizations. The data security and privacy are the most common issues in all the sectors including personal as well as business. With the rise of digital services, data protection and security has become the most critical key areas to respect and safeguard privacy. Data that can be private or public but it needs a secure environment throughout. Organizations that provide digital services must have defined and key processes which include protection regarding the privacy of every individual. In this paper, various types of data security, the risks faced in data security, solutions granted for the problems using technologies, data masking and data mining techniques have been discussed. In addition, this paper also enlightens the recent challenges in data security which will be helpful for novice researchers.

Cyberbullying is upgraded and digitized form of criminal intimidation with various forms and types even in cyber world. The menace of cyberbullying has metamorphosed not just into a technology-based crime but has its magnificent impact on one’s digital as well as physical life. To have insight into various legal and technological aspects of it, research has been conducted through collecting primary data from an online and physical survey / questionnaire. Role of AI, limitations of law and technology in response to control and check cyberbullying have been scrutinized in this paper.

Intrusion detection systems (IDSs) for computer networks play a crucial role in an organization’s performance. IDSs have been created and put into use over the years, utilizing a variety of methodologies to make sure that business networks are safe, dependable, and accessible. In this study, we concentrate on IDSs created by machine learning methods. IDSs based on machine learning (ML) techniques are proficient and reliable at spotting network assaults. However, as the data spaces increase, the effectiveness of these systems declines. Implementing a suitable removing features strategy that can eliminate some characteristics that have little bearing on categorization is essential. To examine the best characteristics in the data, this research suggested an efficient hybrid model that improves computation time and malware detection. This method addresses the problem of high negative result performance and low negative predictive value. Pre-processed data must first be correlated using the Gain Ratio and Co-Relation. Combining these approaches enables learning based on such an essential set of attributes and demonstrates improvement in accuracy and amount of temporal complexity.

Security in ever changing environment is the need of the hour and no business or establishment can survive without a proper and suitable security mechanism. There have been several mechanisms for attending the needs of different sectors. In recent years Zero Trust Architecture has emerged as one of the promising security mechanisms to ensure the utmost privacy and setback resistant authentication mechanism. In this research paper we have tried to propose an advanced level of authentication mechanism known as the stranger trust mechanism.

A VANET is a collection of wireless vehicle nodes that may connect with one another without the need of fixed infrastructure or centralized management. Vehicular ad hoc networks (VANETs) function in a dynamic and unpredictably changing environment that brings numerous potential security risks. One of the types of attacks that affect VANETs the most is the Denial of Service (DoS) attack. Additionally, VANETs can't be secured by following the conventional approaches to protecting wired or wireless networks because of the ever changing network topology. Because preventative methods are insufficient, using an intrusion detection system (IDS) is crucial to the VANET's defense. In this paper, a new intrusion detection system has been proposed by using an Artificial Neural Network-Fitness function based Genetic Algorithm and Support Vector Machines (SVM) for vehicular ad hoc networks to detect the denial-of-service attack.

Digitalization contains a diverse set of information that plays a vital role in investigations, from a forensic stand point. Forensic investigations in the IoT/digital paradigms will need to develop and mature in order to meet the characteristics of IoT. In this paper the authors have investigated digital forensics and have addressed the method of processing the sources of evidences in the digital/IoT ecosystem. The authors have further outlined the guidelines, procedures, current trends and a few criteria and roadblocks in the process of digital investigation. This study lays open ways for investigating strategies and methodologies which support the implementation of digital forensics in the dynamic digital systems. This provides a deep and detailed comprehensive understanding of digital forensics and IoT forensics. Further, current problems and issues are highlighted which will inspire and motivate researchers for further research.

This paper proposes a cloud-based occlusion-aware intrusion detection system that detects the faces of intruders and matches them to a database for recognition. A novel face detection or recognition system is proposed that enables robust face detection and recognition even in the case of occlusion, blurred images, face masks, side views, or even partial views of faces. It does so by focusing on all the visible features like eyes, ears, nose, etc., and using those for facial recognition. The proposed recognition system can match the side view of a face with the front view using the learned embeddings. The user gets an email notification as soon as an intruder is detected whose face does not match any person in the database. The entire system is deployed as a service on Google Cloud. Cloud deployment helps in removing any local computation requirement. The model is constantly updated via the Vertex AI pipelines feature of Google Cloud.

The login mechanism of the web base application currently uses the MD5 hash method used for the encryption of the password. The current state has a weakness of Collision Attack, which means the same hash value will be generated for two or more different inputs. If these values are somehow known, then it can be a threat to the user’s privacy, data and even steal of user’s access. In this situation, as a remedy, we are upgrading our encryption method to the SHA512 method. The collection of data is done by using sources like articles, research papers, journals, patients, magazines, and other online and offline sources. The research is also further divided into parts namely analysis, system vulnerability, and remedy. The report must contain a graphic, pictorial representation by using flowcharts and graphs to make concepts and theory clear about the working mechanism. For testing the finalized product, we conduct a test called UAT. In this test, users come forward to test the application before launching it into the market. In our case, UAT results show that 86% of voters strongly agree with replacing MD5 with SHA51. So, the implementation of patch security at the time of the login process is going to be implemented by SHA512 for further.

To enhance Web Services interoperability for better composition, Semantic Web Services (SWS) architecture offers an opportunity to add higher semantic levels in the existing frameworks using ontologies. Semantically described services will provide better service discovery and allow easier composition and interoperation. Knowledge graphs (KG) use ontologies as the core model to represent formal semantics in knowledge representations and therefore can be effectively utilized in the composition of SWS. Reasoning over knowledge graphs is the new area of research to infer consistent and reliable information from existing data. In this paper, we have proposed and implemented a framework for reasoning over KG using subclass inference that has achieved an average precision of 79.87% and an average query response time of 2.02 s for 37 user queries from 9 domains in the OWL-S dataset.

The massive boost in the technology industry proves as a boon for data science researchers. In the past few decades, new emerging advancements in the field of ML, Deep learning, and image pre-processing makes it possible to automate the task of different fields which required humans rather expert humans’ involvement. Since the advent of neural networking, there has been an enormous demand for acquiring information and generating datasets to support various research projects and to train module work for supporting artificial learning, but due to the lack of diversity in datasets, there are major hurdles during data preparation. There is a need to develop a system that could provide the visual dataset for this communication learning for machine algorithms. To overcome this bottleneck of the neural network, GAN (Generative Adversarial Networks) were created to acquire fake data with the objective of anonymizing users’ information to generate a huge stack of data representations. But working on GAN is not an easy task because it requires a deep understanding of deep learning and image pre-processing. In this paper, a dataset is included to support work for the deaf community which uses sign language to communicate. The task is to read hand signals by detecting the movements and visualizing the communication mode through movements of fingers and to achieve that, personal photos have been captured which are provided as input samples to GAN based model. The various requirements of future work in the same field will require a vast amount of hand sign data which is lacking in the current domain as there have been limitations with different backgrounds, shortage of volunteers, and different gestures to portray generalized reading and storage of such data in a variable format of hand signs have been raised as an issue by many research workers. This data exists in mere thousands and scrapped way over the digitalized media so a methodology is provided to resolve this problem through a tool that will create as many amounts of images datasets as required and, this will provide a healthier feed for machine training that will support future as well as current data requirements for work going on for the deaf community.

Biometric qualities have attracted a lot of study interest over the last few decades. Some traits, such as face and fingerprint, as well as iris, are in recent years the most widely investigated biometric traits in a variety of applications. However, as newer strategies for imitating such features emerge, modalities that are invulnerable to stealth or spoofing attacks are required. This allowed a new biometric trait, the electrocardiogram (ECG), to gain momentum, which is linked with medical diagnosis and is highly resistant to attacks due to its hidden nature and inherent liveness information. But Unimodal biometric systems are attached with some demerits such as Noise in Data, variation in same class, similarity in different classes and more. To counteract these flaws of unimodal systems (single biometric traits), we need a system that overcomes the limitations of any single model. Therefore, multimodal biometric system by combining different biometrics features is required. Multimodal with ECG as one of the traits is the area we are exploring. As liveness detection of subject is not available for most of the modalities so we also need modality to support liveness of subject. We evaluate and discuss existing works in Multimodal biometrics, as well as their proposed methodologies, datasets. The information gathered is utilised to present advancement of ECG biometrics in a multimodal environment.

In the digital era, nature is an incredible and massive source of motivation for learning hard and complex issues in computer science. Nature is a mother of learning because it shows various dynamic, powerful, complex, and exciting concepts. These algorithms are basically designed for solving various optimization problems. These algorithms always provide optimal results for the problem. In a few decades past, many researchers have introduced a large number of nature-inspired algorithms. Few natures inspired algorithms are more effective and useful comparatively other optimization algorithms to provide optimal results with watermarking techniques such as SVD (Singular Value Decomposition), DWT (Discrete Wavelet Transform), and DCT (Discrete Cosine Transform). This paper presents a short description about optimal digital image watermarking algorithms and applications of nature-inspired algorithms: Particle Swarm Optimization (PSO) and Artificial Bee Colony (ABC) & Firefly (FA). The primary purpose of the review is to acknowledge a comprehensive analysis of different nature-inspired algorithms based on their source of inspiration, characteristics, fundamentals, and implementations where these algorithms are properly executed to obtain the optimal solutions for a digital watermarking strategy for digital medical images.

This study focuses to examine the employ of ensemble methods for improving generalization performance which is the most important in the recent study of machine learning. Ensemble learning combines numerous models and deep learning (DL) models with multi-layered ones and also outperforms deep or classic classification models in terms of performance. Deep ensemble learning methods combine the merits of both ensemble learning and the DL models so that the overall result has better performance. This review analyzes current deep ensemble models and provides a comprehensive overview as Bagging, boosting, and stacking are examples of ensemble models. The negative correlation-based deep ensemble models are explicit or implicit ensembles, homogeneous or heterogeneous ensembles, decision fusion strategies, un-supervised, semi-supervised, reinforcement learning, internet, and multilevel featured ensemble models. Deep ensemble models are used in many fields. The author reviews the recent advancement in ensemble deep learning methods and formulates the research objective at the end of this work and future recommendations.

Mental stress is a major problem these days, especially among teenagers. Mental health problems affect students’ health. They affect many areas of a student's life, affecting the quality of life, academic performance, physical health and negatively affecting relationships with friends and family. These problems can affect future employment, income potential, overall health, and can also have long-term effects on the students. Early detection can help them before they go into depression and offer remedial measures that might help to relieve stress. Unfortunately, there is no real-time technique for automatic, continuous, consistent, and reliable stress detection at an early stage. We provide a new framework for instantaneous stress detection. This framework detects hu-man stress based on facial expressions, facial cues (stress score), and breathing patterns (respiration rate). Facial expressions of anger, disgust, fear, and sadness are all signs of stress. The distance between eyebrows and lip movement can also be used as stress indicators. Data on breathing patterns is critical for mental health analysis because it can detect early signs of stress and depression. Skeletal tracking with an Intel depth-sensing camera is used to collect this data. Finally, a machine learning model is built on the collected multimodal data to accurately predict stress levels in multiple subjects.

Due to the development of latest techniques, wireless technology was invented. Basically, it is used in internet access, video conference, and entertainment of mobile technology. People can use the above techniques at any place and any time by mobile communication. Wireless communication means without wires we can transmit data over a long or short distance. In this paper, we will discuss about 1G, 2G, 3G,4G, and 5G technology advantages, disadvantages, and applications discussed. “G” stands for generation of networks. Recently, we have launched 5G technology in India and working on 6G for future aspects.

In spite of various research that has been conducted in the past but there are some challenges that are still into existence related to balancing of workload in cloud applications. There has been a great need for efficient allocation of resources that is handling all the data center, servers & various virtual machines connected with cloud applications. It is the responsibility of cloud facility providers to confirm high facility delivery in an unavoidable situation. All such type of hosts is overloaded or underloaded based on their execution time and throughput. Task scheduling helps in balancing the load of resources and on the other hand task scheduling adheres to the requirement of service level agreement. SLA parameters such as deadlines are concentrated on the Load Balancing algorithm. This paper proposes algorithm which optimizes cloud resources and improves the balancing of load based on migration, SLA and energy efficiency. Proposed load-balancing algorithm discourses all states and focuses on existing research gaps by focusing on the literature gaps. Task scheduling is mainly concentrating on balancing the load and task scheduling mainly adheres to SLA. SLA is one of the documents offered by the service provider to the user. There are various parameters of load balancing such as deadlines which are discussed in the load balancing algorithm. The key focus of proposed process identifies optimize method of resources and improved Load Balancing based on QoS, priority migration of VMs and resource allocations. This proposed algorithm addressed these issues based on the literature review findings.

In wireless sensor network (WSN), energy usage of each node is a main concern to enhance the network performance. A sensor node in wireless sensor networks (WSN) may directly or indirectly communicate to the base station (BS) through single or multi-hop. In that case, the whole network can elect the cluster head to collect the sense data from the remaining nodes of a network. Aggregation is performed at the end of the head node and combined data send to BS. The whole network can also be divided into tiny sensor networks which groups are known as clusters. When clusters are formed then also apply the cluster head (CH) selection mechanism to elect the CH. Now the whole network is divided as per the role of nodes some are CHs and cluster members (CMs). To enhance the energy efficiency, clustering approach can give direction to the sensor network which applies to various applications. So cluster analysis and energy efficiency both are co-related to improve the network lifetime. This research work is helpful to analyze the cluster formation and usage of energy efficiently. This approach is quad clustering which is used to form a single cluster into four clusters. Later the cluster head selection process takes place to make transmission easy with less usage of energy. This work shows the performance evaluation of clusters and the energy consumption of the network. The comparative phase between single and quad clusters considered the following parameters such as distance, number of nodes distribution, and energy usage.

Artificial Intelligence in games is mainly used to control/generate the behavior or the actions of the Non-Player Character (NPC) to a Human Being. AI is a strong selling point of Commercial Games (Video Games). Since games are associated with entertainment, but there are many serious applications of games/gaming such as in medical field, military field. Many other games such as racing game, shooting games they all have different component of AI. This paper presents the usage of AI in gaming and shows their impact in different fields.

Pulmonary fibrosis is a progressive lung illness, it usually gets worse over time as the disease progresses. Scarring develops in the lungs as a result of this condition over time. As a direct consequence of this, people have trouble breathing. Toxic elements are one of the leading contributors to the development of pulmonary fibrosis. These elements include coal dust, asbestos fibers, silica dust, hard metal dusts, and many others. On the other hand, in the overwhelming majority of instances, the physician is unable to determine the precise reason why this sickness occurs. Idiopathic pulmonary fibrosis is the name given to this ailment because it cannot be attributed to any specific cause. This project's objective is to evaluate the performance of several different machine learning models by making predictions regarding the final forced volume capacity measurements and a confidence value for each patient. On the basis of a CT scan of the patient's lungs, it may be utilized on any computer to make an accurate prognosis of the poor condition of the patient's lungs in relation to their ability to operate. A spirometer, which measures the forced vital capacity of the lungs, is utilized in the assessment of pulmonary function (FVC). In the future, it is hoped that pulmonary fibrosis can be identified at an earlier stage. The paradigm of machine learning is helping to improve the effectiveness of the use of human resources while simultaneously lowering costs associated with the social and medical repercussions of this life-threatening illness.

The paper is about the optimal lung disease prediction model. Basically, in this, multiple lung diseases are detected by training a convolutional neural network. And used a deep learning algorithm as it works well with large number of datasets as various dimensions and features of images as well as textual data can be investigated. Lung infections/diseases can be detected by taking blood tests etc. but it is expensive so the main objective of creating this model is to predict using X-Rays to reduce the cost and time of the conduct. Image classification models and segmentation techniques like VGG-16, ResNet50, etc., can be utilized to detect lung infections and illnesses. The semantic gap between the high-level semantic information that humans perceive and the low-level visual information that imaging technologies collect is the fundamental disadvantage of old approaches. The deep convolutional neural network was developed because of the difficulty in maintaining and querying enormous datasets. The developed model gives an accuracy of 82.6% with the loss of 0.2396. The obtained results show the model performance is good and can be used as a secondary opinion tool by medical experts.

The stock market is the place where buyers and sellers come to buy and sell their stocks to get maximum ROI. Stocks go up and down because of the law of supply and demand. The stock market is nonlinear in nature and prediction of stock market trends is a tedious task because of its property to easily get affected by a lot of parameters such as stock and company-specific news, company profile, public sentiment, global economy, etc. Over many years prediction of the stock market trend is a temping problem for most data scientists. To study the massive data generated by the stock market and to perform fundamentals and technical analysis ML and deep learning techniques are effectively used. The fundamentals analysis is generally based on Earnings before interest, taxes, and amortization (EBITA) sheet, quarterly results of stocks whereas technical analysis can be done on daily basis by observing moving averages, candlestick patterns, etc. With the research, it has been observed that the stock market forthcoming trend is highly correlated with candlestick patterns generated in stock markets. There are 42 different candlestick patterns that can form in a timeframe varying from 1 to 4 days but in this paper, only those candlestick patterns have been included which have a timeframe of 1 day. In this paper, a technical analysis-based model has been proposed which uses 4 different candlestick patterns having a timeframe of 1 day to predict the forthcoming trend and the proposed model got an accuracy of 66%.

Motion multiple object tracker model is the objective of the object detection and automated monitoring of the scene for suspicious activity incidents on an ongoing basis. An algorithm tracks the motion of an object in order to predict or anticipate where it will be in relation to its trajectory. The identification method of moving objects using digital image processing techniques in a video series is known as video sequence object detection. The most common tracking method is video surveillance, which is also used in video control systems, intelligent road systems, intrusion monitoring, and airport safety. The tracker has to perform object matching from one edge to the next to get the desired result in the image area. The Identification of object is mainly carried out by the removal of frontal areas. The proposed methodology outlines the conditions under which a particular method yields the most effective results and indicates the location of detected objects for surveillance purposes. This article aims to develop appropriate image processing and computer vision algorithms involving multiple input video sequences and perturbing detection rate study that measures the forward pixel detection rate of background models for different color contrasts for use in road flow and monitoring. This algorithm uses a method that takes into account the pixel intensities of the image, but the results vary in terms of speed, memory requirements to provide accurate results in these situations, and significantly underestimate performance in all cases. Intelligent Transport Systems (ITS) have become more relevant around the world in today's overcrowded transport network and address persistent issues such as transport mobility and safety, in particular in metropolitan areas. Its main objectives are to improve the safety, transport comfort, and efficiency of road traffic.

The global health system is being devastated by the outburst of COVID-19 pandemic. The outburst caused worldwide lockdown. The study found that usage of face mask in crowded areas greatly minimizes the chance of virus transmission. But many people avoid wearing face mask. Specially if people wear a mask while entering a shopping mall or a place with heavy gathering like railway stations, stadiums etc., the chances of infections can be reduced. So, in the present work, the authors have constructed an IoT and AI-enabled Smart Entrance Monitoring Device with an RFID reader for counting number of people who are entering a crowded place with or without mask. This system will measure body temperature as well. Any commercial places like mall, hotel, stadium, railway station or residential entry can benefit from the recommended strategy. This automatic face mask detection system will be cost-effective and require zero manpower for detecting body temperature.

Internet of Things is considered to be the greatest evolving technology and doing wonders in different fields. The advancements in technology have allowed the design of a small and low-cost design that can be connected to internet and can reduce power consumption. Power consumption is very high as power is required in every field. Home automation can be implemented to reduce the household electricity bill. The clumsiness attitude plus our packed daily routine life that sometimes makes ourselves such in hurry situation that sometimes makes us forgot to switch off the lamps. It will cause the electricity bill to rise sharply. Besides, it is one of the electricity wastages that will lead the earth became an unhealthy one. The system is related to home appliances using NODEMCU. Home appliances can help the user to control the devices at home and develop a good condition of the house area that will prevent any loss and damage to the property of any organization. These smart gadgets help to cut down on electricity waste while also promoting energy efficiency. The different technologies include wired home automation system and wireless home automation system and is implemented in various areas such as residential, commercial, and others. Solar panels are also introduced in model for further cut down of electricity and making the system more energy efficient. The proposed system ensures the conversion of solar into electricity.

An interconnected network of 3D virtual worlds is the metaverse. As the virtual environment enables opportunities for extremely immersive and engaging experiences, interest in the metaverse has been expanding globally. This magical development in the metaverse enables new technologies to be integrated and form more advanced tools for a more realistic user experience. To completely meet the expectations of an immersive metaverse, many building blocks are required, such as edge computing, augmented reality, blockchain, and artificial intelligence. In this paper, we have discussed the technologies and applications of the metaverse. We have presented use cases and challenges that the metaverse is currently dealing with, as well as the important advantages of an interconnected world with the help of the metaverse

Body sensor networking is a system that uses sensors to collect physiological data by placing them wirelessly inside, on top of, or across the user's body. This research paper incorporates the comprehensive study of body sensor networking. Multiple research papers were thoroughly investigated, and knowledge is compiled in one. Along with this, the research paper portraits the necessity of body sensor networking for cardiovascular diseases. The aim of this research paper is to display the growing need of body sensor networking specifically for cardiovascular patients and propose a heart vitals monitoring and alert system using Global System for Mobile communication (GSM) technology, trying to generalize the practice of body sensor networking for health monitoring.

Applications that collect data in various forms can add data to the Hadoop stack by partnering with Name Node through an API capability. Using the entirety of the stockpiling and handling force of bunch servers and running dispersed processes on colossal volumes of information are simplified by Hadoop. The Hadoop building blocks can be utilized as an establishment for the making of a few administrations and applications.

The Dutch National Flag problem—DNF Problem, a Computational Problem proposed by the Dutch Computer Scientist—Edsger Wybe Dijkstra—brought a new edge to the domains of Sorting. In this work, we would extend Dijkstra’s Problem to a generalized version, where we would be considering flags with variable numbers of colours and will try to devise an algorithm to arrange them in a monotonic order by its magnitude.

Searching algorithms deal with searching for key $$\mathcal {K}$$ in a heap of data. In this work, an indigenous Searching Algorithm—“The Indian Search Algorithm” has been proposed that will be efficient enough to search for a key in the order $$\begin{aligned} \textrm{log}_{\left( k+\frac{1}{k+\frac{1}{k+\frac{1}{k+\frac{1}{k+\frac{1}{\ddots }}}}}\right) }{\left( n-1\right) }+\textrm{log}_2{\left( n\right) } \end{aligned}$$ where $$\left( k+\frac{1}{k+\frac{1}{k+\frac{1}{k+\frac{1}{k+\frac{1}{\ddots }}}}}\right) $$ is the kth metallic ratio. The Search Algorithm would be an efficient one at least for an Ordered List. In the paper, we have tried to incorporate two children from the same hierarchy—with the parent being the Generic Search with kth metallic ratio, which are Indo-Pellian Search, following the Pell Series, and Indo-Fibonaccian Search following the Fibonacci Series. The Computational Complexity of both have been evaluated.

The coronavirus disease spread has had a lasting global economic impact as companies and governments try to pay for testing and containment procedures. These are necessary steps to limit and reduce the hazard to people’s lives and to mitigate any danger of long-term consequences on economies. The latest epidemic caused a scare among Boeing and Airbus where many people couldn’t travel to high-risk countries, which affects their clientele. The authors use ARIMA model for forecasting the employment rate changes and mobility changes in this research. They start by selecting economic variables (covariates) that may be linked to consumer expenditure and firm revenue. Then, in order to eliminate uncorrelated or slackly correlated covariates, they perform a dimensionality reduction procedure utilizing Gibbs sampling.

Newly discovered coronavirus disease (Covid-19) is an infectious disease related to respiratory illness. Here a texture-based investigation of the coronavirus is performed by analyzing Ultra Violet ( $$\mathrm{UV}$$ ) microscopic images. Considering the seriousness of the issue, we have used the “World Health Organization ( $$\mathrm{WHO}$$ )” information and the “Center for Disease Control and Prevention ( $$\mathrm{CDC}$$ )” in this article. $$\mathrm{UV }$$ microscopic image processing is performed scientifically with the famous and existing image classification texture-based technique. Here we have used a grey level-based co-occurrence approach, a second-order statistical-based approach to analyze $$\mathrm{UV}$$ images. The texture features associated with the images are quantified in four different directions 0°, 45°, 90° and 135° and eight different distances, i.e., $$d=1, 2, 3, 4, 5, 6, 7 \mathrm{and} 8$$ . As a result, we have obtained a changing pattern in the $$\mathrm{Covid }19$$ infected area compared with the human bodies’ non-infected regions. Finally, we have proposed a methodology to analyze $$\mathrm{UV}$$ images of $$\mathrm{Covid }19$$ patients through this research. The obtained results assist the medical and scientific community in fighting these global epidemics.