Innovations in Computer Science and Engineering

Cloud computing is considered as one of the renowned computing methods for pooling and providing various computing resources on demand basis. Cloud computing has grabbed its roots in the IT industry and has become a useful choice for small budget business and organizations. As multiple customers are sharing the same cloud, it will have many security challenges such as malicious user attack, user identity management, auditing, resource management, and integrity control. The main aim here is to provide security to the data by protecting it by unauthorized users during the time of information transmission by using different encrypting techniques such as Blowfish algorithm, RSA algorithm, secure hash algorithm 2, and message digest, on the user data in the cloud. In this paper, we present very recent techniques and algorithms proposed by various authors to secure the stored cloud data.

OFDM stands for orthogonal frequency division multiplexing and is a form of multicarrier modulation (MCM) technique which divides the whole bandwidth into huge number of small sub-carriers. After this, each subcarrier is transmitted in parallel to attain increased data rates. Real data implementation of the proposed OFDM model has been simulated for image signal under AWGN channel. BER is calculated under different SNR values. The proposed work is done using MATLAB–Simulink tool.

In recent times, cloud storage and distributed system have risen to great success owing to their easy use and availability to users. These have helped users to gain access to their data anywhere, anytime solely with the help of internet. But all these features come with greater risks and security threats. Traditional data security methods cannot be considered fully reliable given the advancements in cryptanalysis techniques and sophistication of cyber-attacks. Encryption alone is insufficient to ensure data security and integrity. In this paper, we provide an approach to create a secure distributed file system. We propose a two-layer architecture where the data will first involve interaction with the server. Once we ensure authenticity and integrity of the data, the next step involves storing it securely on the cloud which involves interaction of server with the distributed system which will be completely abstract for the user. The first layer makes the use of digital signature for client authentication and data integrity. The second layer will make use of attribute-based distributed storage of data in encrypted form with hashing techniques to remove redundancy in the data stored in the system.

In the present paper, a crash avoiding at cross roads warning scheme laid on car-to-infrastructure messaging using realistic vehicular dataset is proposed. This meant to avoid collision between the vehicles at intersection and, thus, enhance safety at cross roads. The aim of this paper is to develop an advanced collision avoidance model for connected vehicles which integrates network-level and vehicle-level collision risk. We study the performance of the proposed protocol with an aim to understand the factors that affect successful collision avoidance schemes for cross roads. Through this study, we identified the communication distance as an important factor that influences both the network-level and application-level traffic.

The bulk evolution of IoT devices has resulted in easy monitoring of various daily activities in this modern world. But huge invention of such devices has resulted in easy access of information through these devices, leakage of privacy and resulted in system hackings in household and office areas. Due to flaws in the systems of smart devices results is that IoT devices become vulnerable to hackers. In this research paper, we have proposed a framework in which a channel is implemented through which various firewalls are used to locate and detect such intrusions by the hackers.

In this research paper, a detailed analysis is done on effect of learning rate on machine. Subtle changes in learning rate can bring great impact and can simplify complex research methodologies to a great extent. Our main focus will be on trade-off between learning rate and convergence rate toward an optimal solution. In range of this research, it will be made sure that local optimal solution will not be skipped and the most optimal global solution will be achieved. Keeping all other factors like number of iterations, dataset and algorithm static, we would be verifying our results using experimental analysis done with the help of graphical and statistical observations.

The cryptocurrency space is highly volatile, and predictive systems working in this space are still in their infancy phase. The findings made during an extensive literature survey suggest the lack of a balanced approach and the right combination of data sources, which lead to biased feature sets and discriminative results. These have an impact on the accuracy of the models and skew the classification and prediction results. In this paper, we explore a better approach where a combination of sentiment analysis of social media content, contemporary pricing and market volume data is considered to extract a refined feature set. The features extracted from the preprocessing pipeline will then be used to classify and predict future pricing using a neural network model.

This paper’s proposed method known as assessment of diabetic foot abnormalities for normal and abnormal patients. To evaluate the diabetic foot by using filtered output from a contrast adjusted hyperspectral image and selecting the four seeds points to obtain the cropped image by adding the pepper and salt noisy and apply median filtering from noisy input in order to get the smoothen output image. Then, differentiate the output value normal and abnormal patients. Finally, assess the diabetic foot abnormalities by hyperspectral image. In this article, there are only some qualities which is to regulate enlarge the gap of the figure with chart the principles of the key concentration of figure to original ethics. The progress development is established to arrange in partial, to strengthen the noise which may be nearby in the figure. A number of the applications are included in medical field and geosciences field also.

This paper provides an introduction to application of artificial intelligence (AI) could solve the problems in cybersecurity. The physical gadgets and human impedance are insufficient for managing and ensuring the cybercrimes. Criminals use the Internet to pass on numerous cybercrimes. The specialists want help to stop attacks and security breaks and also react on attacks. The main goal of the cybersecurity is to reduce the attacks by using this AI technology. Applications of artificial intelligence in cybersecurity which exist already and some cybersecurity issues can be resolved by using AI strategies and concluded some useful AI applications.

Popularity of web increases day by day, everything associated with daily life of human being all are connected to web and most of the people spending their time over the web through social networking web apps for their purchasing. Web server stores every activity of users in the form of logs, which contain very useful patterns, henceforth web server log analysis is the vital research area. Web log data analysis has primary step is preprocessing which is meant for dimensionality reduction because web log data is hefty in size and need to normalize the data for further cognitive analysis or other data analysis. Bulky data size degrades the performance of data analytic algorithm so there is necessity of an efficient algorithm for preprocessing over web server log data. In this paper, we put emphasis on data preprocessing. We have proposed the use of genetic algorithm for dimension reduction and normalization of input web server log data. Experimental result shows the data preprocessed data produces higher precision value, precision calculated using MATLAB 2016 classification learner tool.

After death, the human body gets destroyed, brain stops working and human eventually loses his/her knowledge of the brain. But this knowledge and information can be preserved and used for thousands of years. Blue brain is the name of the first virtual brain in the world. This technology helps this activity. This article contains information about the blue brain, its needs, blue brain-building strategies, strengths and weaknesses and more. Collect data on the many types of somatic cells. The analog squares measurements were published on a IBM blue-chip central computer, hence the name “Blue Brain.” This usually corresponds to the size of the bee’s brain. It is hoped that simulation of gallium in the rat brain (21 million neurons) is to be performed by 2014. If you receive enough money, a full simulation of the human brain (86 billion neurons) should be performed, here 2023.

Facial recognition is the process of identifying and/or verifying distinct facial characteristics of an individual in an image or a video stream, pre-recorded or in real time. Hand gesture recognition refers to the process of identifying the configuration of an individual’s hand posture in an image and can be used to verify concurrency of it with respect to a pre-set randomized prompt example to use it in a way similar to CAPTCHA. This paper aims to unify these two similar, yet varying image processing technologies in order to be able to provide a multi-factor authentication system for web applications which cannot be tricked using pre-recorded footage. The face would be identified in the image using convolutional neural networks and then be masked from the image for the hand to be isolated by identifying them using the skin tone. Once the hand has been isolated, SIFT points identified would determine the gesture being performed by the subject, followed by matching the face with the requested credentials if approved. The facial recognition would act like a password and the hand gesture recognition as a means to verify the authenticity of the facial credentials provided.

Quality of service is known to be the feature of network to yield the service that is better for the selected traffic over other technologies like IP-routed networks, frame relay and Ethernet that maybe using any of these or other technologies like this. The basic aim of the quality of service is to provide a dedicated bandwidth, controlled jitter and delay, and it also provides good results on loss features. The other main duty of QoS is to make sure that the priority selected traffic does not significantly affect the flow of other and fail other flows. QoS technologies also provide the primary constructing blocks which will be widely efficient for the application in business area, campuses, wide area network areas as well as the service provider network.

The developing functioning of social media in the field of the travel industry has been expanding on rising investigations of different topics lately. This takes a look at reviews and analyzes the study courses specializing in the world of social media in tourism. Through an in-depth literature assessment, this paper identifies with what we realize approximately social media in tourism and recommends a destiny research schedule on this very phenomenon. The paper indicates that the study on social media in tourism continues to be in its very early stages. And therefore, it is vital to inspire comprehensive research into the influence and the effect of social media (as part of tourism management/advertising and marketing approach) on all aspects of the tourism enterprise that consists of neighborhood communities and to illustrate the financial contribution of social media to the tourism industry.

This is a comparative study of clustering techniques, but focused in the area of market segmentation. By understanding the potential benefits of clustering large amounts of data, the work is to relate clustering into the field of competitive marketing. This is achieved by gathering data from the Twitter using necessary tools and then cautiously applying various clustering algorithms. From these algorithms, we are able to build graphs based on the output, and from these graphs, useful information can be advantages from a strategic marketing point of view. From the clustering of Twitter data, it is easy to identify potential social media influencers. For future implementation, one method is that the companies that are struggling to grow in terms of online marketing can benefit from this study, allowing them to identify their own social media influencers, identify social media trends and determine the online social media market segmentation. All of which will provide them an advantage in further promoting their company more effectively and efficiently.

The work implemented here is crime prediction system (CPS). We first created hypothetical datasets samples of major city areas and different crimes taking place and then we used the algorithms to analyze it. We used HTML and CSS along with PHP, while wamp as a Web server to this application. The objective of the proposed work is to analyze and predict the chance of a crime happening using apriori algorithm. In addition, we used decision tree as a searching algorithm and naïve Bayesian classifier to predict about the crime in particular geographical location at a particular point of time. The result of this can be used to raise people’s awareness regarding the dangerous locations and to help agencies to predict future crime in a specific location within a particular time.

Criminal attacks have drastically increased over the years which make its detection increasingly vital. Fraud detection is a technique of identifying fraudulent activities. We intend to apply clustering techniques in order to analyze and detect fraud or crime patterns from a large set of data. By using various clustering techniques, distinct areas or clusters can be generated by mapping crime instances (i.e., by identifying the factors that lead to fraud). These are areas which have high probability of criminal occurrences which are derived based on historical crime records. Thus, with the help of results obtained based on clustering analysis of the crime data, crime trends today can be identified. Crime can be divided into different types such as location-based crimes, theft, murder, kidnap, fraud, etc., and slums, residential areas, commercial areas, etc., are different types of areas where criminal activities may occur. Primary database is collected based on the types of crimes, the location, and the physical description of the suspects and also the time period in which felony has taken place including the other available data relevant to the analysis. The available data is then processed and clustered, thereby revealing a general crime pattern which in turn helps detect frauds.

Cloud computing provided computing resources which were not imagined before by any business as solutions for their operations. Since the last five years, a new paradigm of service model has emerged which is Function as a Service (FaaS). FaaS is also referred to as server-less computing where the stateless function is responsible for computation of data. These functions are deployed on Amazon Web Services (AWS) Lambda along with other services like API Gateway and DynamoDB to efficiently run the application. The data is transferred using HTTP which is a stateless protocol and it uses its GET and POST methods to exchange data. In this paper, we analyze the deployment of an API on a conventional cloud computing model and compare it with the server-less model, taking account of all the API calls that generate costs.

The rise of regional languages for digital marketing is increasing continuously. Majority of Indians prefer marketing in local Languages. India is among the fastest-growing smartphone markets in the world. This provides an immense scope of brand development and growth of digital marketing. In this paper, the impact of three languages: English, Hindi and Bengali on digital marketing has been analyzed. The analysis shows that online websites using vernacular languages are preferred and have more impact on the consumer.

Digital image processing begins with the acquisition of image which is a start point for further analysis. Whenever an imaging camera takes a picture of some object, more often than not, that particular picture is unusable for the intended purposes directly. Noise in an image is a common phenomenon and is affecting the quality of the image hugely as there is significant amount of deviation in the concentration of pixels of the image. Disturbances and variation in image intensity largely contribute to noising in these images. Image filters are very useful in emphasizing edges or the boundaries between objects or parts of objects in images. Filters provide greater support in the visual interpretation of images. This paper reviews types of noises present in images. It also discusses ways for reducing the quantity of noise and thereby increasing the quality of an image. This paper discusses various filtering techniques and finally how to jointly remove Gaussian and salt-and-pepper noise.

Communication is the ultimate of man’s search for conveying his ideas, emotions, and concepts. Dance is one of the media of communication through which dancers share notion of feelings, with the spectators through gestures, i.e., mudra. Gesture recognition propagates a concept without verbal speech or listening, and in dance recognition, the notion is transferred through various dance poses and actions. This activity in a way really paves way to enhance Indian Sign Language. This study focuses to solve the mudra resemblance in Bharatanatyam through a new system developed with image processing and classification technique using histogram of oriented gradient (HOG) feature extraction techniques and support vector machine (SVM) classifier. SVM classifies the features of HOG into mudras as text labels. Popular feature vectors such as scale-invariant feature transform (SIFT), speed up robust feature (SURF), and local binary pattern (LBP) are hardened against HOG for accuracy and speediness, and this innovative proposed concept is useful for online dance learners.

In recent years, the major issue in cloud computing is providing privacy for getting to re-appropriating information put away on cloud. To store also to share information safely, cryptosystem is utilized. In cryptosystem, the client needs to scramble the data before securing data on cloud and after that decode the data to get to it. This errand can need numerous keys for information encode just like information decoding. But the problem arises with data integrity where data can be modified or the files can be replaced without the knowledge of the data owner. So, to ensure the trustworthiness of the record, the information proprietor figures the hash-based message authentication code (HMAC) signature on each encoded document.

Feature descriptors are vitally important in the broad domain of computer vision. In software systems for face recognition, local binary descriptors find wide use as feature descriptors. Because they give more robust results in varying conditions such as pose, lighting and illumination changes. Precision depends on the correctness of representing the relationship in the local neighbourhood of a digital image into small structures. This paper presents the performance analysis of various binary descriptors such as local binary pattern (LBP), local directional pattern (LDP), local directional number pattern (LDNP), angular local directional pattern (ALDP), local optimal-oriented pattern (LOOP), support vector machine (SVM), K-nearest neighbour (KNN) and back propagation neural network (BPNN) are used for emotion classification. The results indicate that ALDP + Polynomial SVM on MUFE, JAFFE and Yale Face databases gives better accuracy with 96.00%, 94.44% and 89.00%, respectively.

In this paper, we have reviewed blockchain technology along with its usage in industrial IoT. Further, the literature survey has been described that elaborates how blockchain is rising nowadays. Further, the number of applications where blockchain technology can be applied along with smart devices has been discussed. In the last section, we have considered the industrial IoT use case with their recent issues and how these issues can be resolved using the blockchain technology. A proposed security framework in IIoT using blockchain technology is presented. Finally, the paper is concluded with some security solutions in industrial IoT using blockchain technology.

Technologies have been vastly developed in all sectors of the society. Especially in medical field, its growth has been at rapid phase. The newly invented medical equipment has currently been playing massive role in saving the lives of many patients if given the proper and timely treatment to them. Among many modern medical equipment, MRI scanning deserves special mention. Using this technology, we can have the detailed images of organs inside the body as well as categorize and identify the stage of disease. Moreover, with the help of MRI, myocardial disease can be categorized and assessed with several conditions. In particular, we can save patient from a critical situation. However, it is difficult to have an accurate prediction of the cardiac disease. Furthermore, the current medical procedures require more time and medical care to accurately diagnose cardiovascular diseases. Under these circumstances, deep learning method can be useful to have a segment clear and accurate cine image in very less time. A deep learning (DL) technique has been proposed to assist the atomization of the cardiac segmentation in cardiac MRI. We have adopted three types of strategies, according to which we firstly optimize the Jaccard distance to accept the adjective function and then implement the residual learning techniques to integrate it into the code. Finally, a fully convolutional neural network (FCNN) was trained to introduce a batch normalization (BN) layer. However, our standard results show for myocardial segmentation that time taken for volume of 128 × 128 × 13 pixels is less than 23 s which is found when the process is done by using 3.1 GHz Intel Core i9 to be volume.

Social media has become a targeted place for hackers and intruders. The problem is that the detection mechanism which we use is not capable of detecting all the click frauds and has not raised the bar to commit click fraud but is very much effective in the long run. Today’s web browsers support a rich variety of web standards in which a click-bot must be implemented to evade the detection mechanism. A click-bot of heavy size will risk itself of being easily detected by the host. This paper will show a brief review of how this system of detecting fraud-ad works and how we can prevent it from happening with us. This paper reviews what different existing techniques can be used in a more effective way and how they function in a given situation and the ways by which we can keep our data safe from these fraudsters.

Image enhancement technique is one of the very challenging issues in the image processing technique. The main motive of the image enhancement process is to work on an image or picture so that the result should be more applicable than original. Historical docs are very important origin of information, but it commonly suffers from the problem of getting degrade. The main aim is to keep all of these documents secure and alive. The purpose is to enhance the quality and visualization of the historical image. In this paper, we have used improved adaptive histogram equalization of the contrast of the image, and we have analyzed the result.

Load balancing in cloud computing has become an important factor in cloud data centers to ensure proper utilization of resources while preserving the quality of service and meeting service-level agreement of the client applications. The paper presents a literature review on various load balancing techniques that have been applied to the cloud data centers. The techniques involve efficient task to resource mapping during provisioning, or dynamically managing resource allotment for tasks in execution, or optimal utilization of link bandwidth of software-defined networks for data flow. A brief summary of the techniques along with a comparative analysis and advantages and limitations of each have been tabulated. Finally, the paper has been concluded with future scope mentioning the essential points to be taken care of while designing a robust load balancing algorithm for better performance, quality of service and availability.

Live migration is a popular technique used in cloud computing to optimally utilize hardware resources dynamically in the system and reduce unnecessary power consumption in cloud data centers. This paper discusses several prominent live migration techniques applied in cloud computing that aim to evenly distribute load in the system or to consolidate inactive servers. Each algorithm attempts to reduce the overhead of migration by either minimizing the number of migrations or decreasing the application downtime or scheduling the migration while taking available link bandwidth into consideration. A brief summary of each, followed by a comparative account in terms of the major features of the algorithms and advantages and shortcomings of each algorithm is tabulated. The paper is finally concluded by mentioning important points to be taken care of while migrating an application from one physical host to the other in order o achieve better performance.

Research and development is the driving force for the technological future of Mercedes-Benz and guaranties first-class products for its cars. The aim of this work includes innovation, future-oriented products and highly efficient development processes. This study involves visualization of Mercedes-Benz engine testing activities to validate various parameters like fuel consumption, vibration and noise, etc., and generate report templates as per requirements which include plotting charts using graphical packages in R. This process is automated using R shiny dashboard.

This paper offers a medical image compression scheme that includes three stages, namely segmentation (modified region growing (MRG) algorithm), image compression (ROI-discrete cosine transform (DCT) and SPHIT encoding methods, non-ROI-discrete wavelet transform (DWT) and merge-based Huffman encoding (MHE)). Subsequently, the filter coefficients of both the DCT and DWT are optimized using improvised steering angle and gear-based ROA (ISG-ROA). Finally, decompression stage takes place by adopting the reverse process of compression with similar optimized coefficients. Here, the filter coefficients are tuned in such a way that the CR has to be minimal. In addition, an algorithmic analysis is carried out for the proposed model and the outcomes are discussed.

Sentiment analysis and machine-learning techniques play an important role in analyzing social media networks datasets. The customers, who have different levels of demographic attributes pouring views, reviews and feedback on various products and services in social media networks everyday life, this enormous data emerged as major source to extract knowledge to take appropriate decision by companies and business organizations. Most of the sentiment analysis processes ignoring various demographic attributes of customers such as sex, age, occupation, income, location, etc. Different levels of demographic attributes of a customer have their own custom purchase preferences. Depending on the sex, customers will have different preferences, habits and taste of purchasing items. The proposed method focused on sex demographic attribute analysis of the customer to yield effective low-level analysis results. The major challenge in the proposed method is identifying the sex (Male/Female) of the customer by using South Indian names. The proposed system implemented using multi-layer perceptron deep learning method and achieved best train and test accuracy results than decision tree, random forest, k-neighbors, support vector machine (SVM), Naive Bayes. The low-level demographic attribute feature extraction analysis enhanced the effectiveness of the sentiment analysis.

Extracting context from natural language conversations has been the focus of applications which communicate with humans. Understanding the meaning and the intent of the user input, and formulating responses based on a contextual analysis mimicking that of an actual person is at the heart of modern-day chatbots and conversational agents. For this purpose, dialogue systems often use context from previous dialogue history. Thus, present-day dialogue systems typically parse over user utterances and sort them into semantic frames. In this paper, a bidirectional RNN with LSTM and a CRF layer on top is used to classify each utterance into its resultant dialogue act. Furthermore, there is a separate bidirectional RNN with LSTM and attention for the purpose of slot tagging. Slot annotations use the inside-outside-beginning (IOB) scheme. Softmax regression is used to determine the intent of the entire conversation. The approach is demonstrated on data from three different domains.

The need of encryption cannot be underestimated in today’s time. From a travel ticket to a movie ticket, almost everything of need is easily available. This comes with the cost of sharing important information, which if gets into wrong hand, will cause havoc. The aim of this work is to secure the data by fusing the extracts of one-time pad (OTP) and DNA cryptography. In the first level, a symmetric-key algorithm of OTP has been used to make the text secure which is a very random technique. In the next level, conventional DNA cryptographic techniques with few variations have been used. The target of implementation would be the storing and passing of vital information of customers like the debit/credit card details including the CVV code or the virtual passcode. Using DNA cryptography, large amount of storage can be made in the server with an efficient speed and power.

The efficiency of a fuzzy logic-based system is catalyzed by the system design. Fuzzy sets generalize classical crisp sets by incorporating concepts of membership for a fuzzy variable. Each fuzzy set is associated with linguistic concepts that are germane to a particular application. This paper presents an approach for evaluating the region of certainty and uncertainty represented in design of fuzzy linguistic variables. Fuzzy Resembler (FuzR) attempts to capture the goodness of a fuzzy system design using a geometric approach; it can be used for evaluating the design of fuzzy membership space. FuzR is the ratio between region of certainty to region of uncertainty. From the results, it can be inferred that FuzR presents meaningful observations of a fuzzy variable, characterized by trapezoidal, triangular, and gaussian membership functions. FuzR can be used as a design evaluation parameter for evolving fuzzy systems. Knowledge engineers can use it to optimize design of fuzzy systems in the absence of domain experts. Moreover, the level of abstraction provided by FuzR makes it an intuitive design parameter. The significance of this work lies more in its point-of-view than voracious results; the theory and formulation are still young and much more is yet to be conceptualized and tested.

Facial expression recognition (FER) has been one of the actively research topics due to its wide range of application. FER is a very challenging task because of less training datasets. The result of facial expression is the well-classified loss function based on the robust prior knowledge at the end-to-end neural network architecture. The proposed methodology is able to address the task of facial expression recognition and aim to classify images of faces into five discrete emotion categories (happy, sad, angry, neutral, and surprise). Result of this paper is compared with the multiple training datasets and return the maximum appeared face emotion and with highest accuracy. The efficiency a well as the effectiveness of the proposed methodology is more accurate.

In the current scenario, grad students often experience difficultyChakrabarty, Navoneel in choosing a proper institution for pursuing masters based on their academic performances. Although there are many consultancy services and Web applications suggesting students, institutions in which they are most likely to get admitted. But, not always the decisions are staunch since there are different kinds of students with different portfolios and performancesChowdhury, Siddhartha in their academic careers and institution selection is done on the basis of historical admissions’ data. This study aims to analyze a student’s academic achievements as well as university ratingRana, Srinibas and give the probability of getting admission in that university, as output. The gradient boosting regressor model is deployed, which accomplished a $${R^2}$$-score of 0.84 eventually surpassing the performance of the state-of-the-art model. In addition to $${R^2}$$-score, other performance error metrics like mean absolute error, mean square error, and root mean square error are computed and showcased.

Internet of things is a fast-growing field of industry as a lot of IoT applications are introduced in the market. Its popularity is also growing because of its ease of use and its applications serving in vast domains. The demand for IoT deployment over a large scale is increasing at a swift pace. A lot of issues and challenges have worked on, and a lot remains to be explored and solved. It has become a focus of research. A considerable number of researches have contributed a lot in the field of IoT. However, still it is lagging in the security domain, so is the reason for concern for many researchers. There are few other concerns in IoT which are related to the security goals, requirements, challenges, and issues. This paper provides an overview of IoT along with presenting various privacy and security issues in it. The paper also shows how the security in IoT differs from the security in conventional systems. Various lightweight encryption techniques for IoT, limitations of IoT, IoT technologies, and architectures are discussed and compared in the paper.

Web usage mining is used to analyse the user/customer behaviour which is required for business intelligence (BI). The usage of e-payment applications through electronic devices has become more important in organisations and is growing with unprecedented pace. Discovering web usage patterns can result in making strategic decisions for business growth. Especially organisations that need ground truth for exploiting/influencing the customer behaviour. Many researchers contributed towards web usage mining. However, working on real-world data sets provides more useful outcomes. Based on this, we proposed a framework with an EPUD algorithm to perform web usage mining. We have collected electronic payment indicators from RBI dataset and converted it into synthesised server logs suitable for web usage mining. Our algorithm mines the server logs discovers the electronic payment usage and our experimental results reveal the trends in identifying the behaviour of customers in using e-payment systems. The insights in this paper help in understanding the patterns of electronic payment usage for different payment indicators.

This paper presents classification of an image as shirt, T-shirt or trouser for a specific objective by training a convolutional neural network (CNN). The classifier implemented is a significant component of the assistive instrument developed to help people with dementia become more independent with dressing. The work presented in this paper brings out tuning the hyperparameters of the CNN used in the system. A dataset was prepared for the three classes of clothing by capturing the images, pre-processing and labelling the images. Data augmentation was performed on a subset of the original dataset to reduce the overfitting problem. A standard architecture was chosen with convolution, max-pooling and dropout filters which help in dimension reduction, thus enabling faster training of the model. Upon evaluation of the model on the testing dataset, an accuracy of 93.31% was achieved. In order to describe the performance of the classification model, a confusion matrix was plotted.

Learning management system (LMS) provides educational, training and development courses online. It enables the conference, management and display of course content making it easier for all sizes and types of businesses to manage course content. The major share of the LMSs today knuckles down on the corporate market. LMS constitute the greatest wedge of the learning system market. This paper aims to learn the various usability, implementation and adaptation frequencies and the obstacles and facilitators within the LMS domain, specifically in workplaces, by means of comparing a few research articles. The first section tends to introduce six LMS software. The second section comprises of detailed review of LMS platforms. The third section narrows down the results of our comparative review work in a tabular form. The fourth and fifth section gives a graphical representation of our results, concludes the paper and discusses the future scope of LMS.

In this work, we present the efficient detection of multiple actions occurring simultaneously in streaming video of various real-world applications using a frame differencing-based method for background detection. We compare our method with other modeling methods (such as multi-channel nonlinear SVM) for multiple action detection on various video datasets. We demonstrate through quantitative performance evaluation metrics such as performance accuracy, standard deviation and detection F-score, and the efficacy of the proposed method over those reported in the literature.

With the current advances in medical sciences, it is easy to observe the changes happening in the brain in real-time. But the procedure involved is costly and invasive in nature. So, the commonly used procedure is to obtain computed tomography (CT) scans of the brain, which provides static greyscale images. The biggest drawback of a CT scan is that the images are in greyscale; therefore, it is difficult for the naked eye to distinguish the subtle changes in the brain tissues. A wrong prognosis, in this case, could lead to the death of a patient. In this paper, we propose an AI-assisted diagnosis method where a predictive model is deployed, which can discern even the subtlest of the differences in the brain tissues and can help determine any anomalies. The model was trained and tested using CT scans of a rat’s brain, which is affected by Cerebral Oedema (a certain type of disease which leads to accumulation of fluid in the intracellular or the extracellular spaces of the brain). To improve the accuracy of the model, a colour gamut transformation is also proposed. The results after testing the model, with and without the transformation, are tabulated.

Steganalysis recognizes the manifestation of a hidden message in an artefact. In this paper, the analysis is done statistically, by extracting features that shows a change during an embedding. Machine-learning approach is employed here by using a classifier to identify the stego image and cover image. SVM is used as a classifier and a comparative study is done by using steganographic schemes from spatial plus transform domain. The two steganographic schemes are LSB matching and F5 Six unlike kernel functions, four diverse samplings are used for classification. In this paper, the percentage embedding is kept as low as 10%.

Firewall is one of the crucial pillars ofLikhar, Praveen network security. Conventional network firewalls are IP visible and hence vulnerable to network-based attack. IP visible firewalls are IP reachable by attackers from untrusted external network as well as from trusted internal network. A grave situationShankar Yadav, Ravi would result if an attacker managed to break into the firewall and reconfigure it. In this case, attacker can reconfigure the firewall to allow either some specific network service access or in worst case make entire private network reachable by anyone. The risks are Brobdingnagian, once the firewall is compromised, leads to fall the whole network within the mercy of the attacker. To address the security concern due to IP visibility, we designed a stealth packet filtering firewall leveraging the bridging and Netfilter framework of Linux kernel. This paper describes our approach of stealth firewall to overcome limitations of conventional gateway firewall.

Query optimization system proposes an answer-driven approach to information access. Most of the query optimization system aims for information retrieval required by natural language queries. Queries are generally asked within a context, and answers are provided within that specific context. RDF is a general proposition language for the Web, joining data from diverse resources. SPARQL, a query language for RDF, can join data from different databanks, as well as papers, inference engines, or anything else that may reveal its expertise as a guided classified chart. Because of lack of proper architectural circulation, the existing SPARQL-to-SQL translation techniques have actually trimmed a lot of restrictions that decrease their toughness, effectiveness, and reliability. These constraints include the generation of ineffective or perhaps incorrect SQL inquiries, lack of official history, and bad applications. This paper recommended a structure which made use of by an ontology-based moderator system to provide the well-defined semantical design, which (i) supplies a distinct SPARQL semantics used to rewrite the question in SQL; (ii) ontology-based expertise is created for rapid accessibility as well as equate question revising SPARQL to SQL for reliable information retrieval in semantic Internet data of big dataset; (iii) hybrid query optimization framework is proposed for query handling technique for the effective access of customized details on the semantic Internet making use of bundled ontology expertise and also inference engine.

In distributed networks, a single process is selected as the coordinator for each task to be performed. This coordinator process acts as the leader and synchronizes all the processes to execute a particular task. Hence, the leader is first elected before initiating the execution of the task. Two standard algorithms used for leader election are bully and ring election algorithms. But both these algorithms have their respective pitfalls. This paper focuses on combining these two algorithms with the help of hypergraphs to overcome these drawbacks. Initially, the two algorithms are discussed in detail, and their drawbacks are discussed. Further, the concept of hypergraphs is discussed to understand how the two algorithms can be combined, thus introducing a more efficient leader election algorithm.

Perceived quality and value are very essential attributes in the context of brand management. These attributes are traditionally measured using primary surveys. In this work, we propose a methodology to estimate perceived quality and value from online consumer reviews using aspect-based sentiment analysis. We crawled reviews of five popular mobile brands from a reputed e-commerce website. We have applied state-of-the-art text pre-processing techniques to clean the text and to extract the aspects using a semi-automatic approach using dependency parser. The aspects are categorized into five clusters in relevance with benefits consumers get from the brand. Lastly, we have applied TOPSIS, a multi-criterion decision-making algorithm, to rank the brands based on perceived quality scores.

IoT technology has not only revolutionized almost every industry but also it has created a huge impact on our day-to-day lives. The toy industry has also got influenced from this current technology. This paper aims at the development of IoT-based smart toy car which has certain unique features like Peltier AC, Bluetooth audio speaker, obstacle detection, turning-on of indicators on turns, etc. The toy car stops automatically whenever any obstacle is found. Mini AC turns on automatically when the temperature goes above 20°. It has also Bluetooth controlled audio player which can be paired with the Bluetooth of the car on your phone for playing songs or controlling volume. Basically, the proposed system contains different car functions which can be controlled with Internet from anywhere in the world. The system is cost-effective as it uses low power sensors.

IoT has revolutionized the whole world by having mechanism to not only monitor but also control vital statistical information in our environment with the help of variety of sensors. The collected data is transmitted wirelessly to cloud which further processes, stores, transforms and analyzes the data in some usable form (Bahga and Madisetti in Internet of things—a hands on approach. Orient Black Swan, 2015 [1]). The collected information can be accessed through mobile or web applications. IoT-based logger can achieve online real-time monitoring of equipment working status and environmental condition as well as controlling. It can make sure that the equipment works in its security zone and under environment protection state as well as in energy-saving mode. It can not only monitor but one can also analyze the causes of the fault and identify faulty components from the log file created by it on data base server. Using the log file, one can also predict when the environment condition and equipment working status reach at an unacceptable level, and the working of that equipment should be shut down or bypass for maintenance. Thus, the IoT-based data logger (IoTDL) ensures the reliable operation of equipment, personal security and suitable environmental condition. It can also extend equipment life and also reduce equipment failure rate.

The mobile agent devices must be part of any static or dynamic network structure for making the complete network and application functioning. In case of the static networks, the application host distributes specific security protocols for securing the mobile nodes. However, in case of the dynamic networks, the client nodes are directed to select the host nodes for making the application operational. During this process, the host can determine the connectivity of the mobile client nodes based on various authentication schemes. Nevertheless, the mobile nodes cannot justify the authenticity of the hosts. Hence, a good amount of changes persists to get connected to a malicious host node for any client nodes. During this kind of situations, the malicious host node can not only tamper the data, but also de-structure the complete network. Henceforth, this work proposes a novel throughput securing and optimization strategy to protect the network throughput during malicious host connections. This work also proposes malicious host connection fault detection in mobile agents using 2ACK plan. In this paper, we have implanted some security concerns with 2ACK to verify secrecy of the message by confirming first hash code with the objective hash code created. Nevertheless, due to the nature of battery power devices in the network for mobility device agents and WSN, the power consumption of the routing protocols also to be considered for the betterment of the routing algorithms. Henceforth, the energy efficiency became the most important criteria for performance evaluation to be considered. Thus, this work evaluates the performance of routing algorithms for highly dense mobility device agent’s architecture. The major outcome of this work is to propose a novel safe host-based algorithm with lesser energy consumption and evaluates improvements over the existing systems.

The world around us is becoming increasingly complex every day and changes dynamically. The problems that we face require adaptive and scalable systems that can offer solutions with ever-rising level of autonomy. Traditional approaches are becoming obsolete because they were designed for a simpler world. Therefore, any advancement in understanding and solving complex problems can have an impact on the entire set of disciplines in engineering, biology, sociology, etc. In this paper the ant colony optimization (ACO), genetic algorithm is evaluated and compares their performance with the novel proposed adaptive honey bee protocol (AHBeeP). The algorithms, stimulated by the supportive behavior of nature in colonies of animals and social insects, were initially applied to solve the traditional optimization problems. In today’s scenario, the main challenge is to transfer the packets of data from source system to destination system. In the proposed approach, the optimization is used for transferring the data packets based on the honey bees intelligence to communicate each other in the form of dancing language that can be useful for finding the shortest route in the wireless networks and also in optimized way of pathfinding.

The ball-plate system is an unstable open-loop and nonlinear system having 2 degrees of freedom (DOF). This project aims to balance the ball on the plate at the centre or predefined coordinate by adjusting the angle of horizontal plate. The angle of horizontal plate is changed by tilting the plate both in X and Y axes. The actuation is done by two DC servo motors placed in both axes. The feedback of the ball is taken through an overhead camera by processing the images. OpenCV libraries are used for image processing purposes, and Robot Operating System is used as the middleware for the communication between the nodes.

Automatic text summarization is one of the mentionable research areas of natural language processing. The amount of data is increasing rapidly, and the necessity of understanding the gist of any text is just a mandatory tool, nowadays. The area of text summarization has been developing since many years. Mentionable research has been already done through extractive summarization approach; in other side, abstractive summarization approach is the way to summarize any text as like human. Machine will be able to provide a new type of summarization, where the understanding of given summary may found as like as human-generated summary. Several research developments have already been done for abstractive summarization in English language. This paper shows a necessary method—“text generation” in context of Bengali abstractive text summarization development. Text generation helps the machine to understand the pattern of human-written text and then produce the output as is human-written text. A basis recurrent neural network (RNN) has been applied for this text generation approach. The most applicable and successful RNN—long short-term memory (LSTM)—has been applied. Contextual tokens have been used for the better sequence prediction. The proposed method has been developed in the context of making it useable for further development of abstractive text summarization.

Humans have five basic senses that are essentially outward bound. These senses take care of the human survival process. Concurrently, there is an enormous degree of involuntary activity unfolding within a human body that humans are literally not aware of. Some of the examples include water circulating through the body, blood flow within the body, the functional activity of lungs, heart, kidneys, brain, etc. The research proposal discloses a first-of-its-kind method of developing an inward bound ‘sense’ for harnessing perception beyond the ‘five senses’ which are distinctively outward bound. The proposal involves monitoring, recording and analyzing the acoustic fingerprint of the blood around the blood vessels as it flows through them. This would allow the medical fraternity to regularly track and monitor internal bodily parameters which could detect abnormalities in the human body at a very early stage. Thus, the developed ‘inward sense’ would prove to be helpful for diagnosis of any disease in its preliminary phases.

In this current world where everyday people are generating a large amount of data and different business organizations are becoming more and more dependent on it, it has become very important to come out of the traditional methods of data analysis and focusing on the techniques that can prepare much more accurate and valid result to make business decision more easy and simple. This thesis proposes a technique to collect and analyze Twitter posts based on a different keyword-based product searching to generate products market statistical report. Using this program, it can be determined that if any product is getting popularity or losing its market. Few types of results were generated in this project. Each of them has their own type of importance. Overall this type of application can be trusted support for a business analyst or decision makers.

In India, with the advancing urbanization and rapid developments in the transportation has led to a serious concern called Air Pollution. It is becoming an Invisible Killer. Air pollution levels, particularly in cities, are the most alarming threats posed to humanity. However, the existing air quality monitoring systems do not measure the pollutants at the ground level. Although the actual exposure to human beings happens at the ground level, as the emissions from the vehicles are directly inhaled. So, there is a deep mismatch between the ambient levels of air quality measured and the actual pollutants that people inhale at the ground level. This paper focuses to monitor the real-time pollutants using the sensors for the pollutants PM2.5, NO2 and CO as these are the major pollutants from the vehicular emissions and pose serious impacts on human health. Our proposed system uses deep learning-based Long Short-Term Memory (LSTM) algorithm for forecasting the pollutants as this will influence the decision making to improve the city’s quality of air and helps the people plan their day accordingly and take precautions when the pollution levels are unsatisfactory. Finally, our work gives the comparison between prediction of the pollutants at the ground level and ambient air quality levels.

There are a number of languages accent differential applications that detect the different accents in assorted languages. The studies which have done before most of them are based on the English language and different languages throughout the world. A few researches have been performed in Bangla regional language accent differential applications, which is not conclusive for the system to be able to manage Bangla accented speakers. In this paper, we report regional language accent detection experiments of different types of Bangladesh. We demonstrate a strategy to observe Bangladeshi different accents which exploit Mel frequency cepstral coefficient (MFCC) and recurrent neural network (RNN). Listening from the people of different places in Bangladesh creates an accent differentiation results performed by the speakers. This experimental result shows the adaptation of the people to adapt of the regional languages.

There are several works in Bangla handwritten character recognition. Here a new methodology proposed to recognize the character from continuous Bangla handwritten character. The system’s main components are preprocessing, feature extraction, and recognition. There is a strong possibility that is found in Bangla words, and characters are overlapped. This problem often happens in handwritten texts like a consecutive character appears on another character. When it comes to Bangla characters, segmentation becomes much more difficult. To build an effective OCR system of Bangla handwritten text, recognition of characters is important as much as segmentation of characters. Here the main purpose is creating a system, which takes continuous Bangla handwritten text images as an input and then segments the input texts into its constituent words and finally segments each word into individual characters. In this present study, here we used EkushNet dataset model which includes 50 basic characters, 10 character modifiers, 52 frequently used conjunct characters, and 10 digits. By using our algorithm, we are able to segment 95% words from text and 90% characters from the words. Overall, in this present OCR system here recognition and segmentation of characters from handwritten Bangla texts are effectively dealing with the probable problems.

As cloud computing ends up common, increasingly touchier data are being concentrated into the cloud. For the refuge of information protection, delicate information more often than need not to be mixed before redistributing, which makes viable information use a difficult assignment. The proposed structure gives the protection to information in order. In this structure, cloud service provider (CSP) chooses the best host utilizing wake–sleep algorithm. After server choice, client encrypts their information utilizing homomorphic encryption. But it is lagging to reduce computation time and needs an additional security so we adopt and utilize elliptic curve digital signature algorithm (ECDSA). This additional encryption method gives the high security during transmission of data and also avoids bandwidth, less computation time and finally less storage space attained on both the sides.

In today’s world of technological advancements, technologies such as data science, big data and IoT play a huge role in simplifying human life. Access control is a major concern for any data source. Legitimate access to confidential data is important to ensure privacy, integrity, authenticity and confidentiality. It is further very important to devise access control in situations where data is critically private, such as in the case of health care. Access control policies stored in a centralized storage are prone to security breaches due to single point of failure. Blockchain is a decentralized storage system which is based on peer-to-peer network architecture. It is immutable and helps in recording transactions between two parties in a permanent and verifiable manner. The project concentrates on storing the access control policies of a healthcare data source and the transactions that follows in a blockchain. The results of storage of policies and transactions are shown in Remix IDE using Ethereum blockchain framework.

The goal has been to detect disease in mango trees. This paper compares different approaches to extract color features and check the accuracy and applicability for mango trees. The paper proposes variations which helped in increasing the accuracy of features extracted for mango trees: firstly, a customized method of splitting leaf into layers while doing K-means clustering, and secondly, segmenting the region of interest to blocks to help in applying statistical functions more accurately over a region.

We propose a novel foreground subtraction algorithm to extract the Sky pixels and use transfer learning on these processed images using pre-trained ResNet-50 and DenseNet-161 models to classify them. For this study, the Image2Weather’s 2 category weather dataset having 1500 images, is used. By using transfer learning and training, only the last layer by freezing rest of the layers, we have found that the proposed ResNet-50 and DenseNet-161 architectures achieved the classification accuracies of 94.3% and 92.68%, respectively, on the test images and outperformed the state-of-art methods. These results are extremely good as there is a significant improvement of 6–8% in accuracy over the existing models due to foreground subtraction. The proposed method can be implemented in real-time applications to help farmers plan their cropping cycle and for efficient weather forecast activities.

Grouping data suffers from the curse of dimensionality and similarity functions that use all input features with equal relevance may not be effective and the features should be common to complete data. In this paper, Delaunay triangulation method is used to discover and cluster the subspace similarity-based data and finds the closest neighbors by similarity measures, and the triangulation drawing can be done repetitively over the space and cluster. This method avoids the risk of loss of information in any assumed distributed model, and it is a geometric model finds empty circles without any points, only the corner points of the triangles having related points.

Meta-heuristics optimization algorithm is becoming identically popular from the last two decades, and a lot of proposed work has been employed in this field to solve large number of engineering problems, real-world problems, and all other such kinds of problems those are not easy to solve in deterministic amount of time. Such types of problems are known to be NP-hard, and corresponding constraint variables of objective functions contain continuous values. To solve that kind of problem, randomize algorithms (optimization algorithms) come into account that begin with random solutions. This work gives a brief idea about swarm intelligence optimization algorithm, evolutionary algorithms, physical algorithms, and biologically inspired optimization algorithms with their applications. The outcome of these algorithms is prominent in many applications, data set and engineering problems. Some are described in this article out of them.

A medical error is a preventable effect caused due to an improper or incomplete diagnosis of the patient. According to the survey conducted by the Institute of Medicine (IOM) on an average, 97,000 deaths occur annually found in the report for preventable medical error. Hospitals are searching for methods to diminish these errors caused, and adding to its enhancing the analysis of the patient, also the prior data of diagnosis so to ease the examination process and to minimize the cost in the same time. When we go on to analyze the root cause for these toils faced, we observe a lack of medical staff available to each and every patient for their assistance in any instance, which is assuredly impossible. Thus, we go on to update a piece of the present-day equipment available in hospitals that is ECG scanners; these ECG scanners give an outline of the heart rate of the patient through lines of peaks and dips. In this pulse monitor, we give incorporation of emergency message dispatcher to the respective medical personnel for immediate help and also 24-h data synchronizer for future purposes. The major elements in this upgradation are also very easily available and at a modest price. The government can preferably go on with this proposed idea as recruitment of medical staff can be brought down due to the altering of the number of personnel required for each patient and organizing the whole hospital, which benefits the government not only economically but also socially.

The extensive use of the location-based services in today’s communication world has created tremendous interest. Considering the importance of these services, the demand for applications using location-based services is also growing rapidly. While working with the applications of these services, there are many threats related to the issue of security. Security to the user’s data is to be provided from the unauthorized parties in the network. The main idea lies in preserving the privacy of the user using anonymization techniques. In this paper, a method for improving the location privacy of the user is proposed by the popular K-anonymity technique, and the implementation algorithm is also discussed.

This paper discusses relationship extraction among actors/nodes in the text provided. Given a text data, relationships are extracted using natural language processing and shown in a graph. Proper identification of “Nouns” and “Pronouns” can help to identify the actors/characters in the given text. Any relation usually would be between/among the nouns, so this paper first aims to extract “Nouns” and “Pronouns” in the given text and discusses how relation is extracted between/among them. Paper discusses step by step procedure of finding relation among actors/characters in the given plot/text. There are two prominent phases; one is, extracting “Nouns” and “Pronouns,” and the other is the relation among the characters/actors. This methodology is applied on movie story plot and tested. The above said methodology is applied on movie “Baahubali,” as it contains many characters and relations.

Ensemble-based data stream classification process is the most active research work in strengthening the efficacy of ensemble-based data stream classification process. This research is carried out in two different dimensions. First is focused on devising novel ensemble-based data stream classification algorithms to enrich data stream classification task. The second dimension is focused on formulating novel frameworks which propose the novel strategies to aggregate the results of the off-the-shelf classification algorithms. The proposed research work expounds a novel aggregated weighted ensemble framework that aggregates the results of off-the-shelf classification algorithms for data stream classification. The architecture and working principles of the proposed framework, and its role in confronting concept drifts in the data stream classification task are experimentally investigated. Theoretical justifications and empirical evaluation are made on the proposed framework, and accentuate the competency of the proposed framework in terms of accuracy.