Inventive Computation and Information Technologies

Sign language is the primary means of communication used by deaf and dumb people. Learning this language could be perplexing for humans; therefore, it is critical to develop a system that can accurately detect sign language. The fields of deep learning and computer vision with recent advances are used to make an impact in sign language recognition with a fully automated deep learning architecture. This paper presents two models built using two deep learning algorithms; VGG-16 and convolutional neural network (CNN) for recognition and classification of hand gestures. The project aims at analysing the models’ performance quantitatively by optimising accuracy obtained using limited dataset. It aims at designing a system that recognises the hand gestures of American sign language and detects the alphabets. Both the models gave excellent results, VGG-16 being the better. VGG-16 model delivered an accuracy of 99.56% followed by CNN with an accuracy of 99.38%.

Recently, hardware Trojan (HT) is posing a significant challenge to the integrated circuit (IC) industry and has inspired various improvements in the Trojan identification plans. This research study presents the inline assertions for the detection of hardware Trojan at the behavioral level of a system on chip (SoC). In the proposed RTL design, a modified circuit design flow is suggested to incorporate inline assertions into a SoC. Flexible inline assertions are developed in the RTL block within the design module. The router IP design and inline assertions are synthesized and implemented in Xilinx Vivado and Aldec Rivera Pro using Verilog HDL. The universal verification methodology (UVM) is also used to verify the proposed design with the different test case scenarios. The functional coverage and code coverage are analyzed in Aldec Rivera Pro. Parameters such as power and area are analyzed in the Synopsys design compiler (DC).

Nair, Varun Kannimoola, Jinesh M.The dark web is a covered segment of the Internet that provides privacy-protected network access. Tor is a volunteer run prominent dark web network that becomes heaven for criminals to conduct illegal activities. The use of multilayer encryption to achieve anonymity poses a significant hurdle for the law enforcement agency to monitor illicit activities inside the hidden Network. Our study investigates an alternative method to extract the hidden service descriptor from the network. These descriptors also called onion links open a door to hidden services inside dark web. We use a flaw in the v2 protocol to collect the address of hidden service from the memory of a Tor Hidden Service Directory. Automated data extraction and analyzes module provide more insight into contents propagating in Tor network. Using our experiment setup, 4000 onion links are collected and examined. Our analysis shows that socially unjust materials form significant portions of the Tor network.

Geetha Lekshmy, V. Vishnu, P. A. Harikrishnan, P. S.Precision agriculture refers to the application of modern tools and techniques to increase crop productivity in an environment-friendly manner. In the proposed work, a model of self-adaptive system for precision agriculture is developed. This Internet of Things (IoT)-based agriculture system mainly incorporates two functions, automated irrigation and pest detection and is augmented with machine learning models to make it self-adaptive. It handles the sensor failure events automatically by predicting the possible sensor values and keeps the system running without interruption. The system notifies the user about the failure so that it can be replaced later, thus avoiding abrupt termination or malfunctioning of the system. Another adaptive aspect of the proposed system is that it can adjust the system parameters based on prediction of stochastic environmental parameters like rain and temperature. Occurrence of rain is predicted by a machine learning model, and based on this, the system parameters like frequency of getting moisture sensor values are adjusted. This adaptation is fruitful during occurrence of continuous rain when the soil is wet and the moisture content information can be collected less frequently, thus saving the power consumption involved in data collection. The learning models long short-term memory (LSTM) and random forest are used in implementing adaptive functions. The automated irrigation becomes active on fixed times, and the amount of water dispensed is based on the values obtained from soil moisture sensors deployed. The pest detection module captures the images of field and detects mainly the bird pests attacking the crop. The object detection technique, Yolo4, is used to spot the pest.

In recent days, Web design is mainly focused on screen size of the device and for that they are using some predefined cascading style sheet (CSS) like bootstrap. The responsive HTML is making some impact on viewers based on their viewing experience. A few Web sites additionally allow consumers to drag and drop controls to customize the Web page and freely publish the generated Web page. This has limited the user to a set of predefined Web page design styles. Also, the Web design is focusing and realigning the existing and new content of the page according to the new product recommendation by analyzing the previous purchase and similar purchased item. This recommendation is also based on the gender as well. Individual shopper’s decision-making styles are also making an impact on shopping items. The Web details are getting varied on the basis of geographic location and the culture. Depending upon the Web contents, the Web design is changing itself in order to present their contents in a proper way. The researchers are also focusing about the links on the Web page and also about the number of clicks required to reach the particular content and its impact over the user by using this process. The alignment of pictures, videos, and some important text are having ability to contribute for attractive Web design. The papers are based on Web design, and the influence on the user has been taken into account for a survey over here. There are several factors that can influence Web design and the users who use it; we are on our way to developing better Web page designs that improve the viewing experience of users.

In recent days, the value of data stored in disk space (may be localized or cloud) is on the higher side while comparing to the past. In these circumstances, higher the information value provides the higher possibility for data hacking. This situation will develop further and will not be a setback, which implies that the security of those data should be improved as well. This element of security enhancement will be dependent on increasing physical and electronic security. Sensor-based and sensorless approaches can be used to offer electronic data authentication. The sensor-based method depends on specific extra feature on the login device and the state of sensible item with environmental barriers to provide authentication. This raises the question of (i) specified extra feature availability with cost associated with it, (ii) accuracy of sensor devices with respect to sensible item with environmental impact, and (iii) device stability, reliability, and also additional power consumption for sensing device. When coming to sensorless approach for authentication, the simple traditional password scheme is not enough now, and there are some authentication schemes, which will make us to enter different password or pass input ever time, which are already in the pool. This raises the question of the pool size and user remembrance, which is proportional to pool size. If we need a better security, there is a demand to increase the pool size and result in increasing the burden to remember past input for authentication. This research work focuses on reducing the burden of remembering pass input with larger pool. This paper proposes a novel method and implements a bag of password scheme to overcome the aforementioned drawback. As a consequence, with the proposed technique, we are determining the smallest amount of random selection required to select the ideal pool size, resulting in greater authentication security and less complexity from the end user's perspective.

The storage of electronic data as well as the demand for it has become a major problem in today’s society. Data is becoming increasingly centralized in order to provide the flexibility to use it anywhere, at any time, and on any device. Due to the increasing mobility in modern devices, data productivity and accessibility in cloud storage are increasing. Data versatility is expanding on a daily basis, posing a management challenge. All methods of dumping data regularly over a period of time necessitated the deletion and rearrangement of a few data items in order to achieve greater efficiency in the data retrieval process. Currently, the researchers are focusing on the efficient searching algorithms and not on the combined technique of data prioritization, deletion, and rearrangement. The proposed automatic content creation mechanism (AACM) system will create new document after deleting unwanted contents and by merging few existing documents based on the top key words. Each and every document is associated with particular keywords. The proposed system leads to two outputs by considering the text, first to form core points with voting count and then to create new documentary on it. The proposed system can also focus on video, audio, and image in addition to text but however the major focus is given to text, which is the complex one of the four. The mechanism will move from lower priority/older one to higher priority/newer one on the basis of success rate with a particular cluster. This mechanism will save the valid information even from lower priority and older documents. It will also free up the space by deleting the unwanted sentences from older files, and all these depend on the threshold (confidence) value, which is auto-adjusted by the proposed mechanism on the basis of success rate. This will lead to a better memory management and prevention of core historical data.

Voting is a fundamental right given to every citizen of a democratic country, with a minimum age requirement set by the respective countries. As such, one would expect the procedure for voting to be on the cutting edge of technology in terms of security and adhere to the highest standards. This paper proposes and discusses a method of E-voting based on dual-factor authentication in the form of unique identification (UID) number and the fingerprint of the voter for verification purposes. An algorithm for fingerprint recognition is also discussed in the paper along with the efficiency of the algorithm in CPUs of different computing powers. We created a website with the proper focus on securing the personal data of the constituents while also making it legible for the election officials to keep track of the progress of the election and avoid dual/multiple vote casting. The additional security provided by biometric authentication ensures that the system we propose meets the safety standard set by the Information Technology Act, 2000. Based on the experiments and results, we believe that the proposed anti-fraud E-voting system can bring confidence in voters that their vote is secured.

CCTV cameras are found everywhere nowadays and are used to monitor, secure, and protect your property, or at the very least serves as intelligent CCTV footage analysis with sound source separation, object detection and super resolution. However, according to recent statistics, 80% of CCTV footage is discarded in the case of an investigation and is deemed uninformative. The reason being the grainy and low-quality video feed from CCTV cameras. Nonetheless, people thought about it and created video processing software or forensic tools that can improve the quality of the footage. Despite this, the latter are usually expensive or are only available to the authorities. Here developed is an open-source solution that is cross-platform and offers a seamless user interface for your average consumer. The application uses super-resolution to enhance image quality, object detection using YOLO v3, and sound extraction. Using actual CCTV footage as an example, the overall quality and output satisfying results for every functionality.

Nowadays, fall in elders is a major issue almost in all the countries. Sometimes, heavy fall in elders cause serious injuries which leads to major medical care. Fall may lead to disability and also cause mortality to the elderly people. Due to the development of science and technology, the life of the fallen elders is rescued, and the injuries are healed. The newly developed technologies bring happiness and makes the elders life comfortable. At present, fall detection and prevention draws the attention of researchers throughout the world. New technology like Kinect Xbox 360 brings a new way to develop new intelligent system, which could be used to monitor the elderly people in their daily activities. Kinect Xbox 360 is a low-cost device. It tracks the body movements. It is used by the elders doing rehabilitation exercises in the homely environment. Elders who are living alone face the risk of fall. Activity recognition system is a very important technology for elderly people to do their daily activities in their life. Physiotherapy is one of the branches of rehabilitation science which brings differences in the ability and makes the individual to lead a healthy life. In this paper, we are going to analyze various methods of human fall detection and techniques by noticing the daily activities of the elders. We are also going to see different types of machine learning algorithm used for fall detection.

Because of the exponential evolution of digital knowledge, cloud, Internet of things (IoT) technologies, personal information protection and security have gained increased attention. A novel scheme for optical encryption of two-dimensional images is proposed in this paper by integrating image bit swapping strategies in fractional Fourier domains. The secret data hiding process is performed based on LSB replacing technique using secret bits. Image encryption is performed in fractional Fourier transform (FRFT) domain blocks based on the random selection. 3D bit swapping technique is used to perform swapping of randomly selected FRFT blocks. To perform complete encryption, entire blocks will be collapsed based on the randomly generated keys. To evaluate the performance, standard performance measures such as peak signal-to-noise ratio (PSNR) of 41.16, mean square error (MSE) of 0.0056, and correlation coefficient (CC) of 0.99 are presented in various noise conditions. From the performance measures, it is clear that this work achieves better quality when compared to classical techniques.

Johnson, Shanoop Gowtham, R. Nair, Anand R.The growth of malware attacks has been phenomenal in the recent past. The COVID-19 pandemic has contributed to an increase in the dependence of a larger than usual workforce on digital technology. This has forced the anti-malware communities to build better software to mitigate malware attacks by detecting it before they wreak havoc. The key part of protecting a system from a malware attack is to identify whether a given file/software is malicious or not. Ransomware attacks are time-sensitive as they must be stopped before the attack manifests as the damage will be irreversible once the attack reaches a certain stage. Dynamic analysis employs a great many methods to decipher the way ransomware files behave when given a free rein. But, there still exists a risk of exposing the system to malicious code while doing that. Ransomware that can sense the analysis environment will most certainly elude the methods used in dynamic analysis. We propose a static analysis method along with machine learning for classifying the ransomware using opcodes extracted by disassemblers. By selecting the most appropriate feature vectors through the tf-idf feature selection method and tuning the parameters that better represent each class, we can increase the efficiency of the ransomware classification model. The ensemble learning-based model implemented on top of N-gram sequence of static opcode data was found to improve the performance significantly in comparison to RF, SVN, LR, and GBDT models when tested against a dataset consisting of live encrypting ransomware samples that had evasive technique to dodge dynamic malware analysis.

Growing imperviousness and urbanization have increased peak flow magnitude which results in flood events specifically during extreme conditions. Precise and reliable multi-step ahead flood forecasts are beneficial and crucial for decision makers. Present study proposes adaptive neuro-fuzzy inference system (ANFIS) combined with ant colony optimization (ACO) algorithm which optimize model parameters for predicting flood at Matijuri gauge station of Barak River basin, Assam, India. Potential of hybrid flood forecasting model is compared with standalone ANFIS based on quantitative statistical indices such as coefficient of determination (R2), Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). Analysis of results generated by models indicated that ANFIS-ACO model with RMSE = 0.0231, R2 = 0.96014 and MAE = 0.0185 performed better with more accuracy and reliability compared to standalone ANFIS model. Also, results demonstrated ability of proposed optimization algorithm in improving accurateness of conventional ANFIS for flood prediction in selected study site.

Study of consumer behavior analysis within the enterprises is considered as paramount to identify how the customers are satisfied with the enterprise’s services and also predicate how long a customer will exist in the enterprises in future. To achieve better customer satisfaction and to establish a sustainable relationship with the customers, the need for consumer analysis must be performed out expertly. To perform customer analysis in a better way, NB an ML model is studied and analyzed. But due to uncertainties present in the dataset like redundant, irrelevant, missing, and noisy variables makes the NB classifier to analyze wisely. Also violation of independence assumption between the variables in the dataset causes the NB to execute the customer analysis ineffectively. To improve customer analysis with these datasets and to strengthen the NB prediction, this research aims to use of variable selection approach. The variable selection methodology picks the best optimal variable subset by using some evaluation and search strategies to obviate the associated and unrelated variables in learning set and makes the NB assumption satisfied and enhance NB prediction in customer analysis. Three different variable selection methodology is applied in this research (filter, wrapper and hybrid) In filter seven different approaches—Information gain, Symmetrical uncertainty, Correlation attribute evaluation (CAE), OneR, Chi-square, Gain ratio, and ReliefF are applied and in wrapper five approaches—SFS, SBS, Genetic, PSO and Bestfirst are applied and in Hybrid approach combines both filter and wrapper approach. These three methodology works independently to selects the optimal variable subset and uses these variable subsets to enhance the NB prediction in customer analysis data. The experiment performed reveals NB using variable selection methodology gets better prediction compared to NB without variable approach. Also compare to the three variable selection methodology, the Hybrid approach gets better prediction to compare to the other two approaches.

We are proposing a personalized summarization model, which generates an abstractive summary of a random review based on the preference of a specific user. The summary will account the user’s preference on different aspects present in the review. We put forward a Personalized Key Information Guided Network (PKIGN) that pools both extractive and abstractive methods for summary generation. Specifically, keywords present in the review are extracted which are specific to that user, and these keywords are used as key information representation to guide the process of generating summaries. Additionally, Pointer-Guide mechanism is employed for obtaining long-term value for decoding. We evaluate our model on a new Trip-Advisor hotel review dataset, comprising of 140,874 reviews from 41,600 users. Combining the results from both human evaluation and quantitative analysis, it is seen that our model achieves better performance than existing models on personalized review summarization in case of hotel reviews.

The Internet of Things (IoT) is a concept that includes physical devices, web-enabled devices, and the entire network of connections that they use to communicate. The IoT enables these objects to be sensed and controlled remotely across existing network infrastructure, creating opportunities for more direct integration between the physical world and computer-based systems. Nowadays, 5G (fifth generation) systems have the potential to make the IoT concept becomes reality. However, it is difficult for the 5G network against eavesdropping due to the characteristics of 5G networks. In this paper, we have proposed the ElGamal Cryptography with Public Key Infrastructure (PKI) techniques on the communication from the Base Station (BS) to Relay Station (RS) and RS to RS, then finally from RS to Subscriber Station (SS) to against replay attacks, man-in-the-middle (MITM) attack and denial-of-service (DoS). Through the discussion on the proposed mechanism and it has the probability to ensure the confidentiality, integrity, availability, and non-repudiation of the transmitted data. We have discussed the performance analysis of the proposed mechanism and given a conclusion on the discussion. From the result, it shows the mechanism can enhance the security level on the 5G networks.

With the growing elderly population and the importance of a seamless infant care system, wearable tags are vital for continually monitoring health conditions, researching behaviours, detecting events such as falls, tracking location, and so on. With the advancement of wearable technology, several researchers are developing a solution for establishing a seamless human health monitoring system that can be used both indoors and outdoors. The proposed method has a Convolutional Neural Network (CNN) algorithm for recognizing the human biological signals like Temperature (T), Blood Pressure (BP), ECG (S) and Oxygen level (O) and tracking the location (L) of person. Also, the sensed data gets stored in the cloud for analysing the historical data and predict a future event. is the proposed method has obtained a 99.8% accuracy, less complex and compatible to use over other IoT wearable devices and provide a complete report when the authorized person intends to know the status of health and all by using a mobile phone. The proposed Wearable Tag is required to overcome the impact of present COVID-19 scenario.

Underwater Acoustic Sensor Network (UWASN) is a developing technology for exploring Sub-Sea environment and has innumerable applications like deep-sea data acquisition, overseeing the contamination level, calamity prohibition, aided navigation and diplomatic surveyance applications. Previous works on data transmission in UWASN show that the nodes exhaust energy during data transmission and thereby lifetime of the nodes is reduced. This paper integrates localization with Normalized Advancement Factor (NAF) and Packet Delivery Probability to augment the lifespan of network. The NAF is computed from residual energy, Expected Transmission Count and the link cost. Considerable simulations were carried on for analysing the proposed technique and for comparing its performance with the existing VBF, HH-VBF and TORA techniques. Considering the network with 750 nodes, the proposed technique demonstrates better performance with a Packet Delivery Ratio of 95.473%, energy consumption of 0.429 J and Average End to End Delay of 2.73 s.

For water-related industries, the characteristics of wet spells and intervening dry spells are highly useful. In the face of global climate change and climate-change scenario forecasts, the facts become even more important. The goal of this study is to determine the wet and dry spells that occur throughout the monsoon season in peninsular India. The India Meteorological Department (IMD) observations were made over the course of a hundred days, from October 23 to January 30, 2019, with 334 rainy days and 60 dry days. The IMD data provides ten observational characteristics in peninsular India, including maximum, minimum, and average temperatures, rainfall wind speed, atmospheric pressure, illumination, visibility, relative cloud density, and relative humidity. Four statistical factors, such as mean, variance, skewness, and kurtosis, further decrease these characteristics. The observed characteristics and their statistical parameters follow a nonlinear trend, as seen by histogram plots. For assessing the classification performance, a collection of four algorithms is used: Logistic regression, gradient boosting, Gaussian mixture model, and firefly with Gaussian mixture model. During both the dry and rainy spells of monsoon observation, all of the classifiers achieve greater than 85% classification accuracy (average).

Gupta, Kalpesh Dineshan, Aathira Nair, Amrita Ganesh, Jishnu Anjali, T. Sriram, Padmamala Harikrishnan, J.Technology is growing at an exponential rate, and everything around us from shopping to instant messaging and emails, studies, etc. is getting connected to the Internet and becoming smarter. This enabled reconnaissance (or recon) is a collection of procedures and methods, including enumeration, foot-printing, that are used to covertly discover and acquire information about a target system. The protracted recon process requires utmost attention and precision when it comes to handling the device for inspection. There exists a high risk of tampering with the device while inspecting the interior, requiring the replacement of the device. With FCC ID or chip number extraction using optical character recognition, followed by double-checking with the dataset, the specifically designed web scrapers will help to scrape all the information required, including the vulnerabilities from the web, after which a brief report will be generated. Hence, our proposed system automates the process of reconnaissance, saving time and helps in avoiding risks to an extent.

Sudoku is a 9 $$\,\times \,$$ × 9 grid-based puzzle. It is a game where each row, column, and 3 $$\,\times \,$$ × 3 box must have one instance of a number from 1 to 9. In present paper, we shall evaluate three different algorithmic approaches both in serial and parallel configurations that can be utilised to solve a puzzle of Sudoku to assess their comparative performance metrics for differential randomly generated Sudoku datasets. We shall utilise Breadth-first search, Depth-first search, Depth-first search with Breadth-first search parallelisation for sub-tress, for evaluating a large number of randomly generated Sudoku puzzles with a varying number of clues to find the best algorithm based on time and space complexity as well as clue complexity. With this, we shall analyse and develop a best practice algorithm that can be ideally used to solve a large number of puzzles in any given situation in the most time-efficient manner. Our analysis has found that there was a significant improvement in utilising the parallel algorithm over both the Breadth-first and Depth-first search approaches from 28% to over 56%. Even moving from Breadth-first to Depth-first search, we have gauged quite a moderate improvement in performance from 15 to 21%.

A high speed FFT processor is designed supporting 16- to 4096-point FFTs and 12- to 2400-point DFTs for 5G, WLAN. The processor is designed for high speed applications and source code is written in Verilog. Synthesis and simulation is done in Xilinx ISE 14.7. The power dissipation is minimized (20.3 mW) and delay is 9.539 ns and further extension is done using CORDIC processor delay is 7.55 ns. In this paper, high speed VLSI Architecture designed using FFT for 5G Communications. The proposed results are compared with the existed work.

Transfer learning is a technique of training a model for a specific problem and using it as a base for training another related problem. It has been proved to be very effective and has two phases: the pre-training phase (generation of pre-trained models) and the adaptive phase (reuse of pre-trained models). Auto-encoding pre-trained learning model is one type of pre-trained model, which uses the transformer model’s encoder component to perform natural language understanding. This work discusses the bidirectional encoder representations from transformers (BERT) and its variants and relative performances. BERTs are transformer-based models developed for pre-training unlabeled texts, bidirectional, by considering the semantics of texts from both sides of the word being processed. The model implements the above function using two specific functions: masked language modeling (MLM) and next sequence prediction (NSP). The robustly optimized BERT (RoBERTa) variant of BERT with few modifications has significant improvements in removing NSP loss function due to its inefficiency. SpanBERT is another variant that modifies MLM tasks by masking contagious random spans and also uses the span-boundary objective (SBO) loss function. A lite BERT (ALBERT) is another variant with two-parameter reduction techniques: factorized embedding parameterization and cross-layer parameter sharing. It also uses inter-sentence coherence loss instead of NSP. The performance of the BERT’s variants is found to be better than BERT, with few modifications as per the available literature.

The chest X-ray image has been one of the most commonly accessible, which is used for radiological examination whether it be for screening or diagnosis of chest diseases. An enormous amount of study has been done in the field of medical imaging accompanied by different radiologists, but sometimes even for the radiologists it becomes difficult and challenging to examine and review chest radiographs. Our paper aims to provide a new approach for diagnosis of chest diseases into different categories having 15 different labels with the help of transfer learning using pre-trained VGG-16 neural network and localization of the chest images using class activation mapping (CAM). For training the whole model and performing the task, we used a chest X-ray presented by NIH. This has 14 different labels like pneumonia, nodule, and lastly no finding.

This work experimentally evaluated the performance of quantum adders on various IBM quantum hardware. The authors have constructed quantum circuits for one-qubit and two-qubit quantum adders using Quantum Information Science Kit (Qiskit) and run the circuit on seven IBM quantum devices: YorkTown, Ourense, Valencia, Santiago, Athens, Vigo and Melbourne. A detailed experimental analysis of accuracy rate of seven IBM devices are reported in this work. Experimental analysis shows that IBM Athens (5 qubits) provides the best accuracy results (73.7%) in comparison to high-qubit IBM Melbourne (15 qubits) for one-qubit quantum adder. Experimental analysis shows that IBM Melbourne (15 qubits), the only real IBM quantum hardware presently available with qubits higher than 5 qubits, provides an accuracy of 12.8125% over the ideal simulator.

The COVID-19 pandemic is a global emergency that badly impacted the economies of various countries. COVID-19 hits India when the growth rate of the country was at the lowest in the last ten years. To semantically analyze the impact of this pandemic on the economy, it is curial to have an ontology. CIDO ontology is a well-standardized ontology that is specially designed to assess the impact of coronavirus disease and utilize its results for future decision forecasting for the government, industry experts, and professionals in the field of various domains like research, medical advancement, technical innovative adoptions, and so on. However, this ontology does not analyze the impact of the COVID-19 pandemic on the Indian banking sector. On the other side, COVID-19-IBO ontology has been developed to analyze the impact of the COVID-19 pandemic on the Indian banking sector, but this ontology does not reflect complete information of COVID-19 data. Resultantly, users cannot get all the relevant information about the COVID-19 and its impact on the Indian economy. This article aims to extend the CIDO ontology to show the impact of COVID-19 on the Indian economy sector by reusing the concepts from other data sources. We also provide a simplified schema matching approach that detects the overlapping information among the ontologies. The experimental analysis proves that the proposed approach has reasonable results.

Phishing is a type of attack in which attackers obtain personal information such as usernames, passwords, credit card information, and network credentials. They deceive victims by impersonating a reputable individual or entity and conducting specific acts, such as clicking on a harmful connection or attachment or intentionally revealing sensitive information over the phone or through email. In general, phishing sites attempt to deceive victims by pretending they are on a legitimate website to steal their account credentials and other sensitive information. In this paper, we implemented a safe authentication system using secret-key sharing and QR codes. This authentication system has a dedicated mobile application for authentication, which will eliminate the process of entering the website’s credentials and as a result, it will provide robustness for phishing.

In these present circumstances, the future for differently abled students is a big question mark. As the education is turning entirely toward online in which the differently abled students are the most affected ones because their principal way of learning was physical, i.e., using gestures. In this present scenario of pan-epidemic siege, the value of time cannot be ignored for the students who are progressive citizens for a better future. During this intricate time, there is a need to sustain the pace of education for every child and the most important for the differently abled children who are always more enthusiastic in taking on the challenges of life. We at this time, pledge to do our best for the rightful e-learning. In the era of technology, providing education on digital platforms, our idea is to provide some assistance in the field of education technology. The idea is to train a model which will help us to identify and classify Indian Sign Language in the most reliable way. In the previously proposed solutions, the user is restricted to have a definite background so as their model could work accurately. In our system, that limitation is withdrawn. The user can be anywhere, and yet our model would perform the most desirable. We are using OpenCV for pre-processing and a machine learning model is used to recognize hand gestures. This model can then be employed in an Android application for greater perks.

It’s important for software practitioners to understand mutual exclusion in different systems because most of the problems in concurrent systems boil down to achieve mutual exclusion. Mutual exclusion for different types of concurrent scenarios-multithreaded, parallel, distributed can be achieved in different ways by the constructs provided by the programming language. For each of the mentioned types, the performance (or behavior) varies in different ways. The performance of mutual exclusion algorithms is measured by mainly six metrics. This paper shows the comparison between the performance of a chosen synchronization algorithm by analyzing the sequence diagrams obtained from Java Interactive Visualization Environment (JIVE), a dynamic analysis framework for Java program visualization. This paper also presents the results and observations obtained after comparing dining philosophers problem in the above mentioned scenarios. The results are based on the metrics - Message complexity, Synchronization delay and Response Time. The analysis is done on low load performance.

Delay tolerant networks are the best and reliable network in state of emergency such as earthquakes by enabling communication without end-to-end connectivity. For creating communication, store-carry-forward technique is used, that means, if connectivity does not exist between nodes, they store the message till connectivity does not exist and then transfer to other nodes. In this study, we proposed a protocol that tried to deliver the message to the destination node by selecting the best intermediate node based on three features: speed of the node, residual energy of the node, and distance between the neighbor nodes. We also tried to minimize the delay and maximize the delivery ratio by increasing the transmission speed. We simulate our proposed protocol on ONE simulator and compared our method with other three best pre-found protocols. The experimental results convey that our protocol has achieved the delivery ratio of 90% and minimized the delay 4600 s.

After the breakout of this worldwide pandemic situation COVID-19, there arises a severe need for protection mechanisms, wearing a face mask being the primary one. The main aim of the project is to detect the presence of a face mask on human faces on real-time live streaming video. The proposed model is developed using MobileNetV2 which is a deep learning algorithm. The architecture takes the image as input, assigns weights to various objects in the image, differentiates one from another and the neural network output which tells us whether there is a mask or not, and the result is given to the Arduino module by using PY serial software. This model gives an accuracy of 99.9%, and it is connected to the servo-motor which is attached to the Arduino and acts as an automatic sensor door present at various public places. The door will be opened or remains closed based on the output value given to Arduino by the mask detection model designed in proposed study. The door opens only when a person is wearing a mask; otherwise, it remains closed.

The objective of video description or dense video captioning task is to generate a description of the video content. The task consists of identifying and describing distinct temporal segments called events. Existing methods utilize relative context to obtain better sentences. In this paper, we propose a hierarchical captioning model which follows encoder-decoder scheme and consists of two LSTMs for sentence generation. The visual and language information are encoded as context using bi-directional alteration of single-stream temporal action proposal network and is utilized in the next stage to produce coherent and contextually aware sentences. The proposed system is tested on ActivityNet captioning dataset and performed relatively better when compared with other existing approaches.

The devices in the Internet of Things (IoT) communicate through the Internet without human intervention. An enormous number of devices and their generated data leads to several challenges such as data processing at appropriate devices, resource discovery, mapping, and provisioning. The proposed work addresses the management of the workload of devices by offering resources through the fog computing paradigm with less cost and energy consumption. Distributed provision solves the problem of multiple requests having similar response time requirements. It categorizes such requests into different swarms and provides the resources through various fog devices existing in several fog colonies. Each swarm gets mapped to one or more fog colonies considering response time, total resource capacity, and distance between them. Fitness value for all the tasks in a swarm is calculated for binding to fog colony using Multi-Objective Particle Swarm Optimization (MOPSO). In each swarm, the existing requests are mapped to suitable fog devices for processing and avoid overloading and under-provision of fog devices. The performance of the proposed model is evaluated in the CloudSim-Plus framework by the varying capacity of fog instances in terms of small, medium, high, and mixed resources set, tasks/cloudlet length, and response time of requests.

The state-of-the-art presently in the network intrusion detection, both in the network-level intrusion detection system and the host-level intrusion detection system, is completely based on the black box model which learns the pattern from knowledge database or from the dataset to the model. Proposed model is to combine the machine learning-based IDS approach and the domain knowledge incorporating method to build efficient and intelligent IDS which can be employed to detect typical intrusion and future intrusion which is not known. The idea behind is to make some data assimilation process in the features of the dataset such that a reduced and a meaningful feature set representation can be fed in to the model so as to construct intelligent generalized model which will be capable of handling unforeseen attack and new different kind of large data within in limited time period. May be with some compromise in the accuracy of the model but with increased generalizability.

Emotions are the undeniable reliance of information in bridging human and machine intercommunication. Machines are able to recommend better when they can comprehend an individual’s emotions. Producing emotions in the users is conventionally recognised as the fundamental goal of movies. Hence, movie recommendations based on one’s emotional trajectory is key as it allows them to map movie recommendations based on their emotional stage. This paper uses color psychology in capturing user emotions. Two criterions including collaborative filtering (CF) and content-based filtering (CBF) are applied in comparing the apparent user interest profile and identifying like minded users with their cross-recommended items. This is achieved using algorithms with the ability of computing the recommended hybrid and detecting prevalent the emotions. As a result, the system seeks efficiency improvement and enhancement of quality of recommendations.

Vehicle Tracking Technique is a practice that associates the utilization of automaton vehicle locality in singular vehicles with software that automatically determines the geographic location of the vehicle and then transmits the data to the end-user. Present-day, GPS technology was assimilated on vehicle tracking systems for discovering the vehicle locale, however different sorts of automatic vehicle locality innovation can likewise be utilized. In this paper, a vehicle tracking system using GPS and NodeMCU is undertaken to enable users to locate their vehicles at any time with ease and in a convenient manner. The proposed system utilizes GPS that practices satellite technology for navigation and will unremittingly provide statistics like longitude, latitude, velocity, distance traveled, etc. of a vehicle from a remote place to NodeMCU,user's phone as well as send this real-time data to IOT Cloud platform.When a user sends an SMS request, the system responds by sending an SMS to the user's mobile phone with the vehicle's location, and the same information is also stored on Thing Speak. Therefore, advances in technologies and the availability of economical open-source hardware systems are setting a new trend in system designing. The proposed system has been designed using both computer software and hardware, results have been obtained which showed accuracy in positioning and fast response to user commands.

Network security has grown more critical over the past few years as the Internet infrastructure evolves and new technologies become available. An intrusion detection system (IDS) is any kind of security software that can detect and alert administrators of unwanted network access attempts. As a consequence, implementing an intrusion detection system (IDS) to protect network systems is essential. It is a kind of defensive technology that is used to keep networks safe from hackers. IDS is beneficial for both identifying successful attacks and monitoring for suspicious behaviour. Intrusion detection systems provide a solid basis for network protection (IDS). The increased usage of social media networks and cloud computing results in the production of a huge amount of data. With the growth of data production, intrusion attacks may take on a variety of forms. This article focuses mostly on machine learning (ML) techniques for intrusion detection and security. Six different classification techniques are used: KNN, Logistic regression (LR), Naive Bayes (NB) classifier, XGB, DT, and Light Gradient Boosting Machine (LGBM).

Free space optical (FSO) technology is a quick-to-deploy and economical way of getting access to the fiber optic network. FSO technology not only offers fiber-quality connections, but it also offers the sector’s cheapest transmission capacity. FSO systems complement legacy network commitments and function in harmony with any protocol, saving significant up-front investments as a completely protocol-independent broadband gateway. An FSO link may be purchased and deployed at a fraction of the cost of installing fiber cable and for roughly half the cost of equivalent microwave/RF wireless systems. With exception of RF wireless technologies, FSO does not need the purchase of expensive spectrum licenses or the fulfillment of additional regulatory criteria. The purpose of this study is to examine the performance of FSO-based optical access networks. Analysis of the performance in detail, with a focus on BER is also described.

In this digital era, social media generates a large quantity of online financial data, which includes a substantial amount of investor sentiment. On the other hand, only technical and fundamental indicators are no longer adequate to forecast the stock price movement. The investors’ sentiments on social media likes, tweets on twitter, comments and post on Facebook as well as other online financial information like online news, google trend, and forum discussion are also affecting the stock price movement. In particular, researchers have gained a lot of interest for analyzing the financial tweets on Twitter and online financial news to study public sentiments. This would be extremely helpful to develop an efficient solution for automating the sentiment analysis of such vast quantities of online financial texts. Henceforth, the proposed sentiment analysis model aims to predict the stock price movement based on the unstructured data like financial tweets on Twitter and news data used, and this research work also introduces Spark NLP-based text preprocessing pipeline to remove noise data and extract the features using the TFDIF by organizing the text in structured format. For sentiment analysis, two library Textblob and Vader are used. Further, the performance comparison has been carried out. The main aim of the proposed sentiment analysis model is to understand the perspective of the writer from a piece of text whether it is positive, negative, or neutral. In an extensive manner, news and tweets about a security will certainly inspire individuals to invest in that company's stocks, and as a result, the company's stock price will increase.

The wireless sensor network involves sensor nodes, communicating protocols, and gateways for interaction with the Internet. Due to limited memory availability in wireless sensor network, the advanced encryption algorithm of securities and authentication protocol is not deployable due to which wireless sensor networks are prone to attacks such as distributed denial of service and distributed denial of service attacks. The intrusion detection and prevention is used to detect, notify malware activities, avoid, and stop them. The proposed system is mainly to detect and prevent the distributed denial of service and denial of service attack in wireless sensor network. In the proposed model, recurrent neural network is taken as a classifier. The model is validated using the ten-fold cross validation in nine is to one repeated iteration and is then tested for making of false positive alerts on data set (WSN-DS). The accuracy of this model is 99.8% with positive fault rate of 0.3%.

Inconsistency and randomness of wind power impose massive challenges to large-scale wind power production. An accurate production of the wind power for the upcoming hours is imperative, in order that accurate planning and scheduling of the wind power production from conventional units can be accomplished. In the present work, we have proposed three intelligent forecasting models using fuzzy logic, artificial neural network (ANN) and adaptive-neuro-fuzzy inference system (ANFIS) approaches. These models can efficiently incorporate the uncertainty and nonlinearity linked with climatic parameters. To implement these models, the forecasting has been done using historical data of various stations. The performance of these intelligent forecasting models are estimated with statistical indicators and observed that the results obtained using ANFIS forecasting model are found quite accurate. Consequently, ANFIS model can be useful for accurate forecasting of wind power and for efficiently utilizing the wind resources.

The facial emotion classification is a crucial task in human behavior analysis. By taking static images, the emotion is identified from the face expression. It is one of the categories in image processing that is utilized in a variety of disciplines, including human and computer interaction. Some resources are projected to perform automatic emotion recognition, which utilizes benchmark datasets. This research work is focused on real-time dataset that is used to identify six human facial emotions that are implemented by using SVM and tree-based classifier. Experimental outcome symbolizes the top most presentation on the SVM radial basis function (RBF) kernel recognition (95.49%) when associated to the tree-based classifier.

Automotive system has the electronic control unit designed to withstand high-transient pulses and overvoltage conditions. The design and development of USB smart charger with protection circuit play an important role in charging application for mobile phones, tablets, and power banks. In this work, buck converter is designed for a wide input voltage of 9–16 V with 5 V, 6 A output and achieved CISPR 25 class 5 limit compliance for conducted emissions. This circuit is protected from overvoltage/current, undervoltage, and reverse polarity condition which frequently occur in automotive systems. This smart charger is designed to compliance with USB type-A and type-C port which has a separate USB controller for retrieving the charging profile from portable device. Design is carried out for dedicated battery charging of 1.2 standard and simulated using TINA software for normal and faulty conditions. The hardware prototype is developed using AEC-Qualified components and functional testing is performed.

Bitcoin has come out as a huge cryptocurrency success, thus changing the digital transaction (Nakamoto in Bitcoin: A Peer-to-Peer Electronic Cash System. Decentralized Business Review, p. 21260, 2008). In the earlier days, even though proof-of-work (PoW) consensus had performance-related issues, it was not considered as a major issue. The transaction processing mechanism is 3.3 to seven transactions per second with smallest 200–250 byte transactions. While this was sufficient at the beginning, the system has been overloaded over the years resulting in low transaction speed and outrageous transaction cost. Massive growth in cryptocurrency results in scalability issues in the blockchain. With more companies trying to shift their existing system to blockchain, it would be difficult to tackle with the existing PoW consensus caused by its scalability issues (Xie et al. in IEEE Network 33:166–173, 2019). The purpose of this research is to fix the scalability issues and increases the transaction speed by applying techniques such as lightning network, plasma cash, and hard/soft forks.

Log monitoring and analysis plays critical role in identifying events and traces to understand system behaviour at that point in time and to ensure predictive, corrective actions if required. This research is centered towards modelling open-source framework meant for real-time and historical log analytics of IT infrastructure of an educational institute consisting of application servers hosted over Internet and Intranet, peripheral firewalls and IoT devices. Modelling such framework has not only enhanced processing speed of real-time and historical logs through stream processing and batch processing, respectively, but also facilitated system administrators with critical security incidents monitoring and analysis in near-real time. It also allowed forensic investigations on indexed historical logs stored after stream processing by using batch processing. The modelled framework provides open-source, efficient, user-friendly, enterprise-ready centralized heterogeneous log analysis platform with fast searching options. Open-source tools like Apache Flume, Apache Kafka, ELK Stack and Apache Spark are used for log ingestion, stream processing, real-time search and analytics and batch processing, respectively, in this work. Arriving at a novel solution to unify big data processing paradigms stream and batch processing for log analytics, we propose an approach that can be extrapolated to a generalized system for log analytics across a large infrastructure generating voluminous heterogeneous logs.

Technological advancements motivated to develop smart mirrors designed with Raspberry Pi. This paper focuses on the application of smart mirror in home automation, notice board for displaying the news, schedule for the day, weather updates, and room temperature. Home automation includes controlling of electric appliances by voice control using the microphone fitted with the smart mirror, and remote access is also possible by means of Adafruit cloud and Blynk app. PIR sensor is used, and hence, whenever there is no motion detected near to the mirror, the screen gets turned off and thereby increasing the power saving capability. Screen casting can be done using which YouTube videos can be casted on the mirror. Gas leakage in the room can be monitored by the smart mirror in which the gas sensor has been integrated and buzzer gives an alert whenever it is crossing the threshold value. MySQL is used for storing the sensor data which can be used for future analysis. With the help of an own server-based management program wireless devices, the communication between the microcontrollers was done successfully. Big data analytics is integrated in the proposed methodology, and the Grafana is used for data visualization and connecting IoT devices. In practice, the benefits of IoT were demonstrated to bring down the barriers and create a pathway to the mainstream adaptation of IoT smart devices.

COVID-19 is an infectious disease caused by a virus known as novel corona virus or severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Since December 2019, the world is dealing with a pandemic as a result of this sickness. In addition to the medical field, technologies such as deep learning and machine learning aid in the fight against COVID-19. With the use of image or textual data, these technologies can anticipate the existence of disease. The suggested method is a hybrid model that predicts COVID-19 along with illness development using both machine learning and deep learning. Patients’ entire medical records (clinical text data) and chest X-rays are regarded significant data for the suggested method. Because this disease mostly affects the respiratory system, X-rays of the chest are utilized to determine how far the sickness has progressed. A logistic regression model is trained with the clinical text data to classify them as COVID or not. Since the classification is binary, logistic regression is an efficient and easy to implement method. Then a VGG-16 model, which is considered as one of the best vision models, is trained by using several chest X-Rays. The trained model is then used to predict the patient’s status and the progression of disease. A GUI is also developed for a user-friendly experience so that user can directly input the data. The overall output of the proposed model includes the COVID-19 status, the percentage of progression and the masked image of chest X-ray. The combined classifications using clinical data and chest X-Ray improve the effective disease progression detection of COVID-19.

Indoor tracking requires precise localization with the use of short-range radio technology. Tracking the position of humans in an indoor environment is accomplished using Ultra Wide Band communication technology to achieve high accuracy. Ultra Wide Band (UWB) assists in positioning a user in an indoor environment. UWB technology-based devices obtain the position and monitor the movements of a human in an indoor environment. Ultra Wide Band (UWB) technology positions a user with x, y coordinates obtained from timespan and frequency of communication. Positioning with UWB technology is implemented with transit time methodology—Time of Flight (ToF) to measure the running time of light between the tag and anchors. UWB based positioning of an object requires 3 fixed nodes (anchors) to implement the trilateration algorithm. The direct line-of-sight between the tag and nodes is required to achieve high accuracy. UWB technology uses the Deccawave DWM1001C module to identify the location of a user. The system locates the position values of a user in x, y coordinates using UWB technology with the DWM1001C module in Matplotlib and draw the actual location with visualizations using Grafana.

In this paper, a single-phase inverter for avionics application is designed. Usually, the load in the aircraft works on high frequency and with a specific power supply. In the proposed work, a single-phase inverter of 115 V, 400 Hz is designed with an input DC power of 28 V. For the proposed system, DC-DC SEPIC converter is used to step up the input voltage to achieve high gain and an H-bridge inverter followed by a filter to get sinusoidal output. To improve power quality, high efficiency and low total harmonic distortion value closed-loop implementation are validated through detailed simulation in MATLAB Simulink and experimentally verified using a prototype hardware model to get high frequency.

With the increasing settlements, advancement in lifestyle and services, smart cities are experiencing many challenges in solid waste management (SWM) services. The smart city development framework comprises Internet of things (IoT) and information and communication technology (ICT) as key technologies to build an efficient solutions for various services. This research has proposed an IoT-based innovative novel framework to deal with the various challenges involved in SWM services. The framework performs real-time monitoring of waste level in the bin and provides the optimize routes for efficient collection of waste. The overall system constitutes three major components, namely smart bin, smart truck and work server. The smart bin design incorporates various sensors such as ultrasonic, temperature, humidity, load cell, proximity and accelerometer along with radio frequency identification (RFID) tag and wireless communication module. The smart truck is equipped with RFID reader, Global Positioning System (GPS) and wireless communication modules. The work server comprises web server, database management system along route optimization and decision-making programs. The implementation of framework at massive level will reduce the manpower effort and operating cost, optimize the resources needed, improve the environmental quality and contribute to decrease in traffic congestion and noise pollution. Additionally, an experiment was performed using developed prototype which comprised the simplified version of proposed framework. It incorporated the ultrasonic sensor with communication devices. The testing was performed to verify the working of the prototype. Additionally, it was checked that the system was able to detect the different levels of waste in the bin.

Blockchain is an emerging technology with many kinds of cryptocurrencies like bitcoin, Ethereum etc. where one can earn rewards by creating new blocks through the process of mining. But how about only few getting all these rewards and trying to get control over the network? Does it pose a threat to blockchain integrity? Of course, yes which is leading to attacks such as 51% attack, selfish mining attack and double spending. Proof of work (PoW) is a protocol used in blockchain to reduce these problems, but it is not sufficiently secure. So, this paper proposes a technique, using history of miners and the history of their transactions that helps us to reduce the chances of these attacks. We have chosen history of transactions in order to find the genuineness of a miner, thus helping us to reduce double spending. Analysis shows that the risk of these attacks can be decreased using this technique.

Mental illness is a condition that affects the behaviour, attitude and mannerisms of a person. They are highly common in these days of isolation due to the on-going pandemic. Almost 450 million people worldwide suffer from some kind of mental illness. Mental health problems do not only affect adults, but also it has significant impact on kids and teenagers. It is totally normal and understandable to experience fear during the time of COVID-19 pandemic. Loneliness, isolation, unhealthy alcohol and substance usage, self-harm or suicidal behaviour are all projected to escalate as new policies and impacts are implemented, especially quarantine and its effects on many people’s usual habits, schedules or livelihoods. Furthermore, psychiatric disorders have become one of the most severe and widespread public health issues. Early diagnosis of mental health issues is critical for further understanding mental health disorders and providing better medical care. Unlike the diagnosis of most health diseases, which is dependent on laboratory testing and measures, psychiatric disorders are usually classified based on a person’s self-report of detailed questionnaires intended to identify specific patterns. The project would use a person’s tweets, a few customized questions and answers, and a few personal data to measure a person’s mental well-being ranking. This initiative would be immensely helpful to anyone who uses social media sites on a regular basis in order to live a stress-free life and diagnose mental health problems before they get too serious.

Anemia disease is one of the prevalent diseases observed across women and children in most of the developing countries. This is caused due to the iron deficiency in human body. Detecting this disease at the early stage can help the medical fraternity to prescribe proper medicines and come up with alternate solutions that can prolong the patient’s initial stage before it enters into critical stage. Due to the non-availability of historical data of the Anemia patients, there is very sparse literature that addresses the problem of detection of this disease. In this paper, a real-time Anemia dataset pertaining to Indian context is considered and due to the imbalance nature of the dataset, SMOTE is employed for balancing. With the help of decision tree rule-based learning method, rules for detecting the Anemia are derived using two datasets original and SMOTE. The efficacy of the rules is evaluated, and reduced rules are selected based on their individual classifying accuracy. In a quest to give a simple human understandable and optimal rule which can be used by any medical fraternity for detecting the presence of Anemia at different stages, we tried to propose the reduced rules-based method which may come handy. The efficacy of the rules is promising and helps in identifying the presence of Anemia at early stage.

This paper describes the development of a 180 nm standard cell library designed for building power efficient digital circuits. Creating an Integrated Circuit (IC) can be very time consuming if high flexibility or low power is demanded. This paper will try to solve this problem by creating own standard cell libraries, which in turn gives less power. Having these libraries makes it possible to map Verilog code to these libraries, using them instead of predefined cell libraries. The modified full swing Gate Diffusion Input (GDI) design style is used for designing the schematics of basic cells in the standard cell library. The benefits of using this low power technique at circuit level helps to reduce the static power dissipation as compare to CMOS logic for the digital circuits. The static power reduction for sequential designs reduces up to 20% as compared to CMOS logic for the different ISCAS sequential circuits.

The implementation of smart technology is used to improve the safety and efficiency in the drainage monitoring system, which decreases the need of man power and transport costs. The proposed system which accentuates on “SMART CITIES” is used to implement a smarter way of sewage management using smart sensors connected through IoT. The objective of this scheme is to develop a smarter way of conventional drainage monitoring system using smart sensors and IoT. Another key intention is to prevent gas poisoning in sewage maintenance work which can be deadly. An alert system is to be established which constantly monitors the flow, level and toxic gas amount of the sewage pipeline. The status of the particular drainage route is constantly monitored and displayed on the webpage. If any possibility of overflow is detected, the sewage is pumped to the alternate drain automatically until manual maintenance can be performed.

The primary goal of this research work is to develop a smartphone application that allows parents to learn about their children’s vaccination information and keep track of their immunization schedule. Parents can log in with their credentials and upload information about their children. Appointments with paediatricians may be scheduled using the mobile application’s real-time scheduling algorithm to get back the confirmation from the doctor. Further, a mobile a vaccination alert will be sent to the parents via the mobile application notification. Doctors can also know the child details and appointment details in their portal and Parents will be informed about their children’s vaccination details from their portal. The developed mobile application is simple and user friendly.

On March of 2020, World Health Organization (WHO) declared COVID-19 as a global pandemic after months of infecting and claiming many victims. There are some ways by which we can safeguard ourselves against the virus and thereby controlling the spread of the virus. They are following proper sanitization by washing hands with soap regularly, wearing masks and following social distancing, while being present in public places. Social distancing refers to maintaining at least 6 feet of distance between other people. But the main problem is that most of the people ignore these rules and hence the spread of the virus can not be controlled. The project uses computer vision in order to ensure that social distancing is being followed properly, thus helping to reduce the number of victims that the virus may claim. Computer vision is a field of computer science that deals with how computers can gather knowledge and learn from images and videos. It is a rapidly growing field of science thanks to the many advancements in technology over the past few years such as increase in processing power of computers and the exponential increase in data being available nowadays. The system works by taking input from CCTV or other similar image source and then processing the input to find out if any people violate the rules of social distancing and if any violations are detected, the system will consist of an alert module which will alert the respective authorities regarding the violation so that they can do the needful.

SQL injection attacks are widely used by the imposters due to its less complexity and high flexibility. The proposed methodology is intended to perform detection and prevention of such malware scripted SQL queries using SVM. The model first trained with various malware strings and then tested with unknown scripts. It also performs prevention of web application from the SQL malware string using string analyzer and dynamic candidate evaluation. The string analyzer is a grammar-based algorithm that locates the context on the string using regular grammar. Dynamic candidate solution is used to dynamically identifies the malware script using review policy network in which it first generate the parse tree of the input query and then it analyze each node of the tree. It also finds the variation of detection time with respect to accuracy. For the prevention system, simplicity calculates ratio of prevented attack queries out of total number of input queries. The accuracy of the model is good and also the fault rate is minimal.

Injection attacks are most often experienced computer security breaches. Among them, SQL injections with Boolean type of queries and cross-site scripts are mostly done. Probabilistic models fail to adapt to most of the new kind of attacks, due to the changing nature of these attacks. This paper proposes a novel technique for filtering SQL Boolean queries, and these queries are capable of bypassing the existing models but were trapped in this new model. The model uses Rabin-Karp algorithm which is based on the concept of string matching. An SQL query passed as user input is evaluated by the proposed query filter, which is designed to separate malicious queries from the normal queries and flag them as malicious. The proposed model of filtering was experimented on a JavaScript-based web application, designed to test the model with a number of Boolean queries. The results were promising with a maximum accuracy of 96%. Moreover, the model has proved to be effective against the Boolean SQL queries which shows that the SQL injection attacks could be prevented using this model.

Online learning is a well-proven application in all walks of the teaching–learning process that depends on wireless mobile technology and Internet connectivity. It is accomplished through mobile handheld devices and wireless networks. It has proved its importance in this COVID pandemic era for distant learning to facilitate online classes for all university students. The first evolution of remote Internet-based learning was termed electronic learning (e-learning) with wired net-connected desktop computers. Mobile learning or m-learning is accomplished with wireless Internet connectivity-enabled laptops or mobile handheld devices. Now, this technology of the teaching–learning process is widely accepted as the education system of pandemic-affected world and is termed online learning. Online learning is a subset of mobile learning over wireless networks using portable electronic devices. Inhibitors caused due to the existing 4G wireless networks, mobile handheld devices, and the outlook of end users are the main factors selected for this particular study. Inhibitors of online learning over 4G networks were identified through an intensive study conducted by discussions with experts, end users, designers and learners. The methodology adopted for this research work is the fuzzy matrix of cross-impact multiplications applied to classification (MICMAC). The fuzzy-MICMAC method is used to analyse the selected inhibitors of online learning over wireless networks. A digraph with all possible interconnections, correlation study and the comparative analysis of fuzzy logic approach to the binary logic ISM-MICMAC method were implemented. More outlined results were achieved with the fuzzy analysis. The driving power-dependence abilities of these inhibitors were recorded, which will resolve many undue variables in the future implementation of the system.

The article formulates an improved mathematical model for clarification of imagery for an area of the earth’s surface. An improved mathematical model for clarification of geospatial information has been formulated in general form, which can be presented as a result of the action of an operator that transforms coordinates, operators that conduct image clustering, and operators that carry out zoning according to some criterion. The co-ordinates transformation operator, clustering operators, zoning operators, and their explicit form are presented. This model, based on the model of forming imagery, performs the inverse transformation of the imagery coordinates into spatial coordinates and clustering imagery into particular classes according to their texture and color. This takes into account geographic zoning. A model for clarification of geospatial information in operator form is obtained.

Facial biometrics continues to be the preferred biometric bench-mark even though other human body signatures are also used. But there are many problems with face recognition. A slight change in the image, which may be due to illumination, or to the environment, can drastically affect the results and accuracy. These changes may not even be noticeable to the human eye, but can affect the face recognition process. There are some advantages, such as contactless, easy to use and, keeping in mind the current situation, reduces the risk of getting infected by touching buttons or by simply waiting in the queue for a long time. In this paper, an effort is made to review different kinds of face recognition methods comprehensively. PCA, LDA, SVM, and various hybrid techniques are used for the face recognition approach. This review examines all of the aforementioned techniques as well as various parameters that pose challenges to face recognition, such as illumination, pose variations, and facial expression.

There is currently no cutting-edge methodology for genre transfer of music as creating music is challenging. CNN-based GANs are used to generate pictures, whereas transformers are used to generate text. Since music does not fall into both categories, none can produce the greatest effects on its own. To address this issue, this research work employs TransGAN, a paradigm that employs transformers (rather than CNN) in the design of GANs. We believe that the proposed approach may be used to provide cutting-edge performance in music generation. A dataset of 64 × 64 and 128 × 128 pixel mel spectrogram images was used, and the model was able to transfer genre and detect musical patterns.

Any IT organization needs an infrastructure consisting mainly of computers with different capacities of hard drives and other components. The operation and maintenance costs of such decentralized networks are high; hence, troubleshooting and security can be difficult. Educational institutions are also forced to cut budgets and provide better solutions for student’s learning and development. The main purpose of this paper is to analyze and develop a system based on the concept of diskless booting a hybrid computing, in which the diskless client PC is devoid of the secondary storage device responsible for storing the necessary boot-oriented and operating system files. This paper also provides critical analysis of how inexpensive, energy-efficient, and fully managed diskless clients have proven to be a better approach which enhances speed and reduces cost, keeping in mind the existing infrastructure.

Video surveillance is being increasingly adopted for ensuring safety and security both in public and private places. Automated prediction of abnormal events like theft, robbery, murder etc., from continuous observation of surveillance videos is a multidisciplinary study involving computer vision, deep learning and artificial intelligence. Deep learning-based video analysis and categorization is a most researched topic. Many deep learning models based on long short-term memory (LSTM) are proposed for automated prediction of abnormal events. This work does a comparative analysis of six LSTM-based deep learning models for abnormal event prediction from surveillance videos. Deep learning models of ResNet, VGG16, VGG19, 3DCNN, Inception V2 and Inception-ResNet-V2 are combined with LSTM for prediction of abnormal event from past observation of events in the video stream. These six models are run against different benchmarked abnormal event detection datasets and performance is compared in terms of accuracy, loss and execution time. It is observed that Inception-ResNet with LSTM provides training accuracy of 80% and test accuracy of 80% compared to other models.

Face detection is an artificial intelligence [AI]-based methodology, where a cascade function is prepared with a bunch of information and input data. It is prepared from a great deal of positive and negative pictures. It is then used to identify objects in different pictures. The proposed model’s classifier essentially increases the precision and strength of robustness search on faces with demeanor variation and vague forms. In the element extraction stage, this research work leverages a philosophy for increasing efficiency through the connection of two strategies: mathematical (geometrical) element-based approach and independent component analysis strategy. To implement face coordination step, this research work proposes a model that combines several neural networks to coordinate with the mathematical highlights of the human face. Since the proposed model connects several neural networks, it is termed as multi-artificial neural network. MIT + CMU information base is utilized for assessing the proposed strategies for face identification and arrangement. At last, the experimented results on Caltech dataset show the possibility of implementing the proposed model.

In any network system, the intrusion is undesirable, and organizations are constantly searching for solutions that could effectively detect intrusion and, consequently, help them to react appropriately. However, packaged enterprise solutions provided by industry-leading companies usually leave little room for optimization and control in the hands of the users and sometimes incur costs that small- and medium-sized organizations want to curtail, especially if the solutions are smart. This work demonstrates how such organizations may build their home-grown Deep Learning-based Intrusion Detection Systems (DL-IDSs) and integrate them into their existing network. We have implemented an Intrusion Detection System for Small networks using deep learning architectures. The proposed system evaluated on UNSW-NB15 dataset including more than 250,000 network packets and has obtained an accuracy of 89% in discriminating between abnormal and normal packets and 74% for various nine network attack types classification.