Volume 5, Issue 1 (2024)

Rise of Identity and Access Management with Microsoft Security
Sheetakshi Shukla, Kirti Jain
Page 1-7

Abstract— Identity and Access Management (IAM) is a pivotal element in modern cybersecurity strategies, enabling organizations to manage user identities and control access to digital resources securely. This paper focuses on Microsoft’s comprehensive suite of IAM solutions, emphasizing the innovative capabilities of Microsoft Entra ID as a central component within its ecosystem. The discussion spans Entra ID’s role in IAM, Multi-Factor Authentication (MFA), Conditional Access policies, and Microsoft Entra Privileged Identity Management (PIM). This research explores the dynamic landscape of IAM in the context of Microsoft security, addressing challenges and opportunities posed by contemporary cybersecurity threats and evolving work environments. Key topics include the integration of IAM solutions with Microsoft 365 services, the impact of remote work on identity governance, and the effective implementation of conditional access policies to enhance security without compromising user experience.
Furthermore, the paper investigates the role of IAM, specifically Microsoft Entra ID, in meeting security and compliance requirements. It delves into data protection, threat intelligence, and compliance reporting within the Entra ID framework. As organizations navigate hybrid environments that span on-premises and cloud infrastructures, the research examines the intricacies of managing user authentication in such diverse setups. The study concludes by emphasizing the importance of adapting IAM strategies continuously to address evolving cybersecurity challenges within Microsoft’s security ecosystem. By referencing the latest Microsoft Entra ID documentation and industry best practices, this research contributes to a deeper understanding of the significance of IAM, specifically Microsoft Entra ID, and its practical implications for organizations seeking robust identity and access management solutions.

Keywords— IAM, Microsoft Entra ID, MFA, Conditional Access, Identity Governance

DOI: 10.5281/zenodo.10621038

⇩  PDF

Security Features in Fingerprint Biometric System
Shilpa, Kirti Jain
Page 9-18

Abstract—At present, embedded systems operate in every environment on the planet. Many complex applications with previously unheard-of capability have been made possible by recent technological advancements. Regardless of the ability to shield critical data from malevolent attacks, security and privacy remained a prevalent concern for these systems. These worries are warranted since horrifying tales about embedded systems are told by the past security lapses and their aftermath. With the development of technology, the attacks are gradually changing and becoming more sophisticated. As a result, fresh approaches to security implementation in embedded systems are needed. This paper uses a case study to illustrate how security features are integrated into fingerprint biometric systems during the requirements analysis stage and maintained throughout the embedded system life cycle. A comparative analysis is provided between different biometric technologies, including face, fingerprint, iris, palm print, hand geometry, gait, signature, and keystroke. In order to provide more precise safety requirements or functions, the goal of this work is to analyze, break down, and convert the risks and countermeasures found during the requirements analysis utilizing the abuse case. Additionally, by examining the system requirements and outlining the primary procedures for biometric system protection in this article, we have demonstrated how security features can be incorporated into the biometric fingerprint system.

Keywords—Information leaks, fingerprints, abuse cases, countermeasures, and threats

DOI: 10.5281/zenodo.10641797

⇩  PDF

Analyzing the Sentiment of Social-Media for Predicting Depression using Supervised Learning and Radial Basis Function
Yogesh Sahu, Pinaki Ghosh
Page 19-23

Abstract Sentiment analysis is a recent phenomenon that involves comprehending individuals’ feelings across many situations in their everyday existence. The use of social media data, including textual information as well as emoticons, emojis, and other visual representations, will be employed throughout the whole of the process, including the analysis and classification operations. Previous study conducted several experiments utilising Binary and Triple Classification methods; however, it has been shown that multi-class classification offers more accurate and precise categorization. The data would be partitioned into many subcategories according to the polarity in a multi-class classification. Supervised Machine Learning Methods will be used throughout the categorization operation. Sentiment levels may be monitored or analysed using social media. This study investigates the use of artificial intelligence techniques to analyse sentiment in social media data with the purpose of understanding or detecting it. The poll used visual campaigns to analyse social media data, which included words, emoticons, and emojis, for the purpose of emotion detection using diverse machine learning methods. The SL-RBF Algorithm demonstrates higher accuracy in sentiment analysis.

Keywords– Sentiment Analysist, Radial Basis Function, Accuracy, Multi Class Classification, Precision.

DOI: 10.5281/zenodo.10658820

⇩  PDF

Analysis of Forward Pass RNN with Hyperbolic Tangent Function for Software Defect Prediction
Swati Rai, Kirti Jain
Page 25-29

Abstract Software failure prediction and proneness have long been considered critical challenges for the IT industry and software professionals. Conventional approaches may detect software defects inside an application, but they need previous knowledge of problems or faulty components. Automated software fault recovery models enable the programme to significantly predict and recover from software issues via the use of machine learning techniques. This feature reduces mistakes, time, and money while also making the programme run more smoothly. A software defect prediction development model was given using machine learning techniques, which could enable the programme to carry out its intended purpose. A range of optimisation evaluation benchmarks, including as accuracy, f1-measure, precision, recall, and specificity, were also used to evaluate the model’s performance. The FPRNN-HTF (Forward Pass RNN with Hyperbolic Tangent Function) deep learning prediction model is based on convolutional neural networks and its hyperbolic tangent functions. The evaluation process showed how well CNN algorithms were used and how accurate they were. Additionally, a comparative metric is used to assess the proposed prediction model in comparison to other approaches. The collected data showed how well the FPRNN-HTF approach performed.

Keywords– FPRNN-HTF (Forward Pass RNN with Hyperbolic Tangent Function), precision, recall, specificity, F1-measure, and accuracy.

DOI: 10.5281/zenodo.10666637

⇩  PDF

Performance Analysis and Diagnosis of Thyroid Disease Detection using Deep Learning
Aparna Anand, Kirti Jain
Page 31-38

Abstract Thyroid Disease is very common problem that affects lot of people around the world. That is why, it is important to diagnose this disease accurately and effectively. But it is not that, this disease has not been diagnosed yet. It is been done by doctors for long time. To facilitate doctors, I developed a deep learning model to identify Thyroid disease accurately. This is accomplished by providing dataset which was provided by Garavan Institute, located in Sydney [1]. The dataset comprises features related to thyroid conditions, where thyroid positivity being labelled as 0 and negativity as 1. This model involves densely connected layers which employs Rectified Linear Unit (ReLU) activation function. Techniques such as dropout and batch normalization are incorporated to enhance generalization and preventing overfitting. The Adam optimizer is utilized for gradient descent. The training process is monitored using metrics such as accuracy and loss, and model’s performance is also evaluated using a separate test set. Standard evaluation metrics such as accuracy, precision, recall and F1_score is computed to access the model’s predictions. We can figure out how good our computer is at recognizing different thyroid conditions with the help of confusion matrices. A classification threshold is explored to convert continuous predictions into binary outcomes. This model involves preprocessing steps such as feature scaling with the help of Min Max Scaler. Data Augmentation is done to enhance the model’s ability. This model includes deep learning techniques such as neural networks, optimization algorithms and evaluation metrics to develop a robust binary classification model for thyroid disease diagnosis.

DOI: 10.5281/zenodo.10666856

⇩  PDF

Comparative Analysis of Microservices Architectures: Evaluating Performance, Scalability, and Maintenance
Mahesh Kumar Bagwani, Gaurav Kumar Shrivastava
Page 39-43

Abstract— Microservices have become a key architectural paradigm in the ever-changing field of web application development. This study compares and contrasts microservices architectures in great detail, paying close attention to each one’s scalability, maintenance, and performance. This research analyses a variety of microservices frameworks and reveals the subtleties of their architecture through a methodical assessment.  Through an examination of critical performance indicators like response times, scalability under different workloads, and ease of ongoing maintenance, the study seeks to identify best practices and draw attention to potential issues related to each architecture. The knowledge gathered from this research will help architects and developers choose or optimize microservices frameworks with confidence. This paper not only contributes to the academic discourse but also offers pragmatic guidance for real-world applications, ensuring that the chosen architecture aligns seamlessly with the specific needs of a project. Embracing a holistic approach, this research provides a nuanced understanding of the trade-offs inherent in diverse microservices approaches, fostering a more robust and informed development community.

Keywords— Microservices, Monolithic, Jenkins, JMeter, KVM

DOI: 10.5281/zenodo.10709588

⇩  PDF

Showcasing Retrieval and Language Models for Information-Rich Natural Language Processing (NLP)
Jitendra Singh Kustwar, Gaurav Kumar Shrivastava, Nikhil Chaurasia
Page 45-48

Abstract— In the realm of Natural Language Processing (NLP), the integration of retrieval and language models has become paramount for handling information-rich content effectively. This paper presents a comprehensive exploration and showcase of advanced techniques in combining retrieval and language models to enhance the capabilities of information-intensive NLP systems. The primary objective is to bridge the gap between knowledge retrieval and contextual understanding, enabling applications to seamlessly navigate extensive knowledge bases. The paper begins by surveying state-of-the-art retrieval models, delving into their strengths and limitations in extracting relevant information from large datasets. Subsequently, it explores the landscape of language models, including transformer-based architectures such as BERT and GPT, focusing on their abilities to capture intricate linguistic nuances and semantic relationships within the context of information-rich tasks. Our approach involves the careful composition of these models, emphasizing the synergy between retrieved knowledge and contextual understanding. The proposed models aim to not only retrieve relevant information but also comprehend and integrate it seamlessly into the context of natural language understanding.

To demonstrate the efficacy of the showcased models, we present practical applications across diverse domains, including healthcare, legal, and scientific literature analysis. We evaluate the models using rigorous metrics, assessing their performance in terms of accuracy, precision, and recall. The language model is constructed using advanced deep learning methods, specifically focusing on recurrent neural networks (RNNs) and transformer architectures. The model is trained on large datasets to learn intricate patterns, semantic structures, and contextual nuances within the language.

Keywords— Natural Language Processing (NLP), Machine Learning (ML), Deep Learning (DL), recurrent neural networks (RNNs).

DOI: 10.5281/zenodo.10711142

⇩  PDF

Malware Identification Using CNN and Deep Forest with Transfer Learning
Nivedita Wahane, Chandan Kumar
Page 49-55

Abstract— Malware, also known as malicious software, is a set of code that performs malevolent operations with the sole purpose of harming or taking advantage of the individual, government, device, service, network, or for monetary gains. Thus, it has become a priority to find ways to detect malware as a step towards identifying and preventing the malware attacks.

Researchers have proposed many different malware detection and classification models using various techniques like static-, dynamic-, visualization-based analysis, Machine Learning, Deep Learning, hybrid (i.e. combining two or more different methods), transfer learning approaches, and more.

In this paper, malware identification methodology is proposed using hybrid deep learning models with transfer learning. After converting the suspected file into grayscale image, the proposed methodology will accomplish the task of Malware Identification (i.e. is the provided input malware or benign?) by using CNN (which will be pre-trained by Transfer Learning) for feature extraction from malicious/ benign file’s image and Deep Forest for classification.

This proposed methodology will happen in the hope of, first, achieving better accuracy by training the Deep Learning models for malware identification using Transfer Learning; second, to give the end user with only the needed information of whether the file is infected or not as the information about malware families would be of no use to the regular users; and third, since training the model for both identification and classification task will only increase the pre-training, computational time, and resource consumption, to counter this, this model is proposed.

Keywords— Malware detection, Visualization-based detection, transfer learning, CNN, Deep Forest

DOI: 10.5281/zenodo.10711191

⇩  PDF

Fake News Detection Using Deep Learning: A Comprehensive Review
Amit Kumar Saxena, Kirti Jain
Page 57-63

Abstract– Organizations from various domains are working to find effective solutions for detecting online-based fake news, which is a major issue at the moment. It can be difficult to recognise fake information on the internet because it is frequently written to deceive individuals. Deep learning-based algorithms are more accurate at detecting fake news than many other machine learning techniques. Previous reviews focused on data mining and machine learning approaches, with little attention paid to deep learning techniques for detecting fake news. Emerging deep learning-based techniques like Attention, Generative Adversarial Networks, and Bidirectional Encoder Representations for Transformers, on the other hand, were not included in earlier surveys. This research looks into advanced and cutting-edge false news detection techniques in depth. We’ll start with the negative consequences of fake news. Then we’ll talk about the dataset that was used in earlier research and the NLP approaches that were used. To divide representative methods into several categories, a complete overview of deep learning-based techniques has been presented. The most often used evaluation measures in the detection of false news are also reviewed. Nonetheless, in future research paths, we propose additional recommendations to improve fake news detection techniques.

Keywords– Natural language processing, machine learning, deep learning, and fake news

DOI: 10.5281/zenodo.10711229

⇩  PDF

Smartphones Capturing Gait Biometrics - A Deep Learning Paradigm
Mohini Parihar, Pinaki Ghosh
Page 65-70

Abstract— This research paper aims to explore the feasibility and effectiveness of utilizing smartphones as a tool for capturing gait biometrics, employing a deep learning paradigm. Gait biometrics, the study of human walking patterns as a unique identifier, holds significant potential for applications in security, healthcare, and personalized technology. Traditional gait recognition systems have faced challenges in terms of accessibility and user-friendliness. In this context, smartphones, being ubiquitous and equipped with various sensors, present a promising avenue for unobtrusive and continuous gait data collection. The paper investigates the role of deep learning techniques in analyzing the gait data obtained from smartphones, aiming to enhance the accuracy and reliability of gait recognition systems. To assess the viability of smartphones as a platform for capturing gait biometrics. To employ deep learning techniques to develop a robust gait recognition model using data collected from smartphones. To compare the performance of the proposed smartphone-based gait recognition model with traditional methods. The significance of this research lies in the potential transformation of gait biometrics from specialized, controlled environments to real-world, everyday scenarios. Smartphones, being an integral part of modern life, offer a convenient means of continuous gait data collection without requiring additional hardware. The application of deep learning in gait analysis enhances the model’s ability to recognize subtle and complex patterns, contributing to improved accuracy and reliability. The findings of this study could pave the way for widespread adoption of gait biometrics, with implications for security systems, healthcare monitoring, and personalized technology interfaces. The fusion of smartphones, gait biometrics, and deep learning stands to revolutionize the landscape of human identification and interaction in various domains.

 Keywords— Gait recognition, inertial sensor, person identification, convolutional neural network, recurrent neural network.

DOI: 10.5281/zenodo.10718864

⇩  PDF

A Review of Recent Studies on Prediction of Cardiovascular Disease
Irfan Khan, Pinaki Ghosh
Page 71-75

Abstract – The heart is a vital organ of the human body. It’s the main part of our circulation system, and cardiovascular disease has been a common cause of mortality in the last few decades. It’s increasing day by day at a rapid rate. So, it is necessary to build a system to diagnose cardiovascular disease beforehand. Machine learning is a branch of artificial intelligence; it learns from historical data, builds prediction models, and, whenever it receives new input data, predicts the outcome. The authors discussed the various machine learning algorithms used to measure the accuracy of cardiovascular disease. The prime contribution of our work is to study the various machine learning techniques used to measure accuracy to predict heart disease.

 Keywords: cardiovascular disease, machine learning, supervised, unsupervised. Logistic regression.

DOI: 10.5281/zenodo.10718879

⇩  PDF

Technological Breakthroughs Shaping Smart Energy Administration in Urban Centres
Pooja Vishwakarma, Pinaki Ghosh
Page 77-81

Abstract—the rapid urbanization and increasing energy demand in recent time have necessitated the advancement of smart cities that can efficiently supervise energy resources while ensuring feasible and well-being for their residents. Effective energy administration is essential to achieving these goals. This article explores the cutting-edge technology innovations driving intelligent energy management in smart cities. This integrates smart power grids, renewable energy sources, sophisticated metering infrastructure, demand driven systems, energy-efficient architecture, and the intricate world of data analysis. By examining the benefits, challenges, and future prospects of these technologies, this paper provides a comprehensive overview of how technology is shaping the energy landscape of modern urban environments

Keywords- Smart City, Power grids, Energy Management, Intelligent Cities, Renewable energy

DOI: 10.5281/zenodo.10718943

⇩  PDF

A Brief Survey on Techniques for Protein Sequence Analysis
Prativesh Pawar, Pinaki Ghosh
Page 83-87

Abstract-There are currently a lot of biological data available, and data mining is essential in sorting the data. Many research on the use of data mining in bioinformatics have been conducted as a result of the efficacy of data mining techniques in all facets of computational biology. Over the past two decades, a body of literature on data mining methods in bioinformatics analysis has grown. A periodic examination of survey articles is essential, and grouping them makes it easier for the researcher to identify the study. This document also teaches non-specialists how to select among a variety of currently used strategies based on their strengths and weaknesses. In this study, an effort is made to offer a thorough analysis of the algorithms that are optimal for obtaining the desired outcome.

Keywords: Deep learning; natural language processing; protein annotation; protein language model; protein sequence embedding; survey of embedding models.

DOI: 10.5281/zenodo.10718952

⇩  PDF

Intrusion Detection System Based on Particle Swarm Optimization in Mobile Ad-hoc Network: A Survey
Shruti Dixit, Navneet Kaur, Shalini Shahay
Page 89-93

Abstract- Mobile Ad-hoc network (MANET) is the assortment of cooperative wireless nodes without existence of any access point or infrastructure. Due to problems like wireless radio, limited battery power, limited bandwidth and dynamic topology environment, nodes are susceptible for intrusion and attack. Security is an important field in this type of network. Each node in a MANET is capable of acting as a router. Routing and routing protocols are important aspects having various security concerns. The bio-inspired approach known as Particle Swarm Optimization (PSO) based on Swarm Intelligence (SI) is suggested for finding solution against attacks in the network. In this paper a survey of different types of attacks are presented and intrusion detection (ID) mechanisms based on PSO is discussed.

Keywords- Mobile Ad-Hoc Network (MANET), Intrusion Detection Systems (IDS), Swarm intelligence (SI), Particle Swarm optimization (PSO)

DOI: 10.5281/zenodo.10718959

⇩  PDF

Autism Spectrum Disorder Detection Using Machine Learning
Priya Gyanchandani, Gourav Shrivastava
Page 95-98

Abstract— autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by a variety of behavioral and social problems that can be effectively managed through intervention and support if diagnosed early. However, early diagnosis of autism spectrum disorders is still very difficult. Current diagnostic methods often involve lengthy and expensive tests, including clinical examinations and interviews, making them impractical for large-scale screening. The aim of this study is to use a noninvasive and cost-effective method to solve important problems in identifying autism spectrum disorders in childhood. This study focuses on the potential of facial features (key features of a person’s face) as an indicator of autism spectrum disorders. These studies highlight the need for a comprehensive, multidisciplinary approach to autism diagnosis that involves clinicians, researchers, data scientists, and the autism movement to improve early identification and support of individuals on the autism spectrum. This article focuses on machine learning for ASD diagnosis. It includes SVM, DT, RF, KNN, clustering and other methods.

Keywords— autism spectrum disorder, machine learning, early detection, SVM, RF, DT

DOI: 10.5281/zenodo.10719317

⇩  PDF

Performance Analysis using Machine Learning for Code Mixed Languages in Sentiment Analysis
Shruti Mathur, Gourav Shrivastava
Page 99-104

Abstract- Social media podiums like Twitter, Facebook, and Instagram have gained a lot of attention these days and have become one of the most prominent platforms to communicate, share thoughts and voice opinions. Detection of human emotions like happiness, sadness, anger, sarcasm etc. in textual communications has, therefore, become very important Sarcasm is a way of communication that creates gap between the anticipated meaning and the genuine meaning comprehended from the conversation. Communication and human relations over social media sites like Facebook, Twitter circles around a lot of sarcasm and debates. Sarcasm detection is an important processing problem which is needed to understand the human and machine communication better. Code mixing, as the name suggests, alludes to blending various dialects or more than one language in a solitary expression or a sentence. For a multilinguistic country like India, code mixing has become a very common practice on social media platforms since the pandemic since it is easier for the users to use their native language along with expressing their feelings. This paper aims to understand the gap between the emotion and the contextual meaning by using different machine learning approaches for Sarcasm Detection of code-mixed Hi-En dataset. The algorithms used in this paper are Bernoulli Naïve Bayes, Logistic Regression and Support Vector Machine. SVM outperforms all the used algorithms giving an accuracy of 87.36%.

Keywords: Code-mixed language; sarcasm detection; Natural language processing.

DOI: 10.5281/zenodo.10719323

⇩  PDF

Image Based Potato Leaf Disease Detection Using CNN-LSTM Model
Gargi Sharma, Gourav Shrivastava
Page 105-110

Abstract— In recent years, the agricultural industry has faced significant challenges in food production due to the prevalence of crop disease. Potato is one of the most well-known crop cultivates in India and diseases such as early blight and late blight, significantly impact the quality and yield of potatoes, and manual interpretation of these leaf disease is time consuming and labour intensive. To address the issue, this paper proposes a novel approach for potato leaf disease detection by combining CNN and LSTM algorithms. In the proposed algorithm CNN is used to extract different features from leaf images and then with the help of LSTM classifier, the result was perceived. The objective of the model is to develop an accurate and efficient model that can identify diseases affecting potato crops and the proposed model has achieved an accuracy of 98.5% on the potato dataset.

Index Terms—Deep learning, CNN, LSTM, leaf Disease

DOI: 10.5281/zenodo.10719315

⇩  PDF

A Review of Deep Learning Mechanisms for Intrusion Detection and Prevention in Network and IOT based Environments
Nikhil Chaurasia, Neeraj Sharma
Page 111-115

Abstract– As the Infrastructure is growing, we found a tremendous growth in Digitization, every enterprise is focusing on premise Data center or on the Rented Cloud, so to meet security prospective from Intrusion is one of the major concerns. A New terminology is being used as CSP’s (cyber-physical systems) instead of Datacenter and with the Evolvement of Deep Learning (DL) Concepts and its efficiency DL procedures finds a great scope to remove all the vulnerabilities by priory identifying Risks and then afterwards by prevention from any king of malware impacts .In our survey we basically focusing the Application of Deep Learning (DL) Procedures to Build a secure systems  by implementing strong Neural Networks (NN) by providing suitable Training with the malicious Data sets and then afterwards to develop a good prevention capabilities, Deep learning is the subset of machine Learning (ML) and also in the previous scenarios ML  techniques proves to be very much stable due to their self-Learning and enhancing Capacity in terms of Weighted Attributes they are used in Spam detection,DoS Attacks, probe Attacks, Host based Attacks, Network based Attacks etc. In our survey paper we will enlist multiple DL Learning Procedures as per the Attack Types.

Keywords: cyber-physical systems (CSP’s), Deep Learning (DL), Machine Learning (ML), Neural Networks.

DOI: 10.5281/zenodo.10779137

⇩  PDF

Performance Analysis of Modified Convolutional Neural Network Model for Recognition of MODI Character Set
Anshika Jain, Maya Ingle
Page 117-121

Abstract— The “MODI Script” is an ancient script that originated in Maharashtra, Western India, and is widely used to create official documents. In the area of handwritten character recognition, a recognition system for MODI might be developed in order to interpret it better. Deep learning algorithms are used in character recognition applications to identify patterns accurately and efficiently. We proposed Modified CNN model for recognizing handwritten MODI character set. Our model comprises of five convolutions, two pooling, and one fully connected layer. Increasing the number of convolutional layers improves recognition accuracy by extracting more features. We trained our model with modified layers using 56-character set of MODI including numerals, vowels, and consonants, along with matras for each consonant. A total of 33,600 images have been used for the training and testing process. The results of the experiment indicate that the proposed model achieved a higher accuracy rate of 99.95%.

Keywords: Character Recognition, MODI Character Set, Modified CNN Model

DOI: 10.5281/zenodo.11186653

⇩  PDF