Volume 10, Issue 3 - IJIRTM
May - June (2026)
Impact Factor: 5.86 | Volume 10 | Issue 4
Performance Analysis of Liver Diseases Classification using Machine Learning Classifier
👥 Neetiraj Singh Lodhi, Dr.Rajneesh Choubey
📙 Abstract : Early detection through supervised learning techniques applied to diverse datasets is crucial in reducing mortality rates. With ongoing advancements in medicine, a significant amount of data has been collected, particularly in healthcare, where extensive data generation occurs. This data is processed and analyzed through data mining applications for knowledge extraction. Mining algorithms effectively predict patient diseases by utilizing appropriate learning strategies. Chronic kidney disease (CKD), hepatitis, cancer, and diabetes represent major global health concerns, making their prediction a significant focus for researchers. This dissertation primarily analyzes various classification algorithms, including Support Vector Machines (SVM), K Nearest Neighbor (KNN), and Extra Tree Classifier, by comparing their performance utilizing liver patient data. The study employs different machine learning classification techniques to diagnose liver conditions at an early stage, yielding performance metrics such as accuracy and other relevant parameters.
🔖 Keywords :️ Accuracy, support vector machines, Supervised machine learning, Random forest.
Detection and Prevention of Distributed Denial of Service (DDoS) Attacks in Cloud environments: A Comprehensive Survey
👥 Rajeev Ranjan, Darshna Rai, Chetan Agrawal
📙 Abstract : Cloud computing is the dominant paradigm for scalable IT services, but its openness and resource-sharing approach make it a target for DDoS attacks. This paper is a comprehensive survey of DDoS attacks targeting cloud infrastructures, the evolution of detection and prevention mechanisms, the role of machine learning and deep learning in automated threat mitigation, and the growing importance of SDN as a defence platform. We categorise 60+ research studies from 2018 to 2025 by approach and effectiveness and critically evaluate their strengths and weaknesses. Our assessment provides a uniform taxonomy, comparison tables, and architectural diagrams to help researchers and practitioners navigate this complicated world. Federated learning, explainable AI, and zero-trust cloud DDoS protection architectures are our final research problems and intriguing future prospects.
🔖 Keywords :️ DDoS attacks, cloud computing, intrusion detection, machine learning, deep learning, SDN, botnet, mitigation, network security, traffic classification.
Educational Data Mining for Student Performance Prediction: A Systematic Review
👥 Mirza Hufeza Baig, Prof.Aashish Kumar Tiwari
📙 Abstract : Educational Data Mining (EDM) has become a rapidly growing interdisciplinary research domain that applies data mining, machine learning, and artificial intelligence techniques to extract valuable insights from educational datasets. Among its various applications, student performance prediction has gained significant attention due to its potential to enhance academic achievement, identify at-risk students, and support informed educational decision-making. This review paper provides a comprehensive analysis of recent advancements in student performance prediction using Educational Data Mining approaches. Various predictive models, including Decision Trees, Naïve Bayes, Artificial Neural Networks (ANN), Support Vector Machines (SVM), Random Forest, Ensemble Learning, and Deep Learning techniques, are critically examined and compared. Furthermore, the study explores the key academic, demographic, and behavioral factors that influence student performance. Current challenges, existing research gaps, and emerging trends in the field are also discussed. The review highlights that hybrid machine learning frameworks, advanced feature selection techniques, and intelligent learning analytics significantly enhance prediction accuracy and contribute to the development of effective educational support systems. The findings of this study provide valuable insights for researchers, educators, and policymakers seeking to improve student success and institutional performance through data-driven educational strategies.
🔖 Keywords :️ Educational Data Mining, Student Performance Prediction, Machine Learning, Learning Analytics, Classification, Artificial Intelligence.
Genetic Algorithm and Neural Network in Educational Data Mining
👥 Mirza Hufeza Baig, Prof.Aashish Kumar Tiwari
📙 Abstract : Educational Data Mining (EDM) has become an effective research area for extracting meaningful knowledge from educational datasets and improving student learning outcomes. Accurate prediction of student performance plays a vital role in identifying academically weak students and supporting educational decision-making. This paper proposes a Behavior-Based Student Classification System (SCS-B) that integrates Genetic Algorithm (GA) and Back Propagation Neural Network (BP-NN) for student performance prediction and classification. Student data are collected through a structured Student Questionnaire (SQ) comprising learning techniques, personal information, student behavior, intellectual factors, and comprehensive abilities. The collected data undergo preprocessing through outlier detection and dimensionality reduction using Singular Value Decomposition (SVD). Genetic Algorithm is employed for optimal feature selection, while BP-NN is used for classification and prediction. The proposed model classifies students into four categories: Class A, Class B, Class C, and Class D based on their academic and behavioral characteristics. Experimental evaluation demonstrates that the proposed SCS-B model achieves superior classification accuracy compared with Support Vector Machine (SVM), Multilayer Perceptron (MLP), and Multi-Tier Student Performance Evaluation Model (MTSPEM). The results confirm that integrating behavioral factors with machine learning techniques significantly enhances student performance prediction and educational decision support.
🔖 Keywords :️ Educational Data Mining, Student Performance Prediction, Genetic Algorithm, Neural Network, Behavior Analysis, Machine Learning, Student Classification.
SEAST: A Simulation-Driven Framework for Enhancing Human Resilience Against Social Engineering Attacks
👥 Jatin, Khushi, Vaishali Wadhwa
📙 Abstract : Social engineering attacks have become one of the most prevalent and damaging threats in the modern cybersecurity landscape, primarily because they exploit human psychology rather than technical system vulnerabilities. Despite continuous advancements in security infrastructure such as firewalls, intrusion detection systems, and endpoint protection solutions, attackers increasingly succeed by manipulating user behaviour through techniques such as phishing, smishing, spear-phishing, and baiting. Industry reports consistently indicate that a significant majority of data breaches involve human error, highlighting the urgent need for cybersecurity defences that focus on strengthening the human element, often referred to as the “human firewall.†Traditional cybersecurity awareness programs largely rely on static training materials, including presentations, videos, and compliance-driven quizzes. While these methods may increase theoretical knowledge, they often fail to produce measurable behavioural change or prepare users for real-world attack scenarios. Moreover, such programs lack personalization, real- time feedback, and meaningful analytics to assess user susceptibility and organizational risk. To address these limitations, this paper presents SEAST (Social Engineering Awareness and Simulation Tool), a next- generation, simulation-driven cybersecurity awareness framework designed to enhance user resilience against social engineering attacks. SEAST integrates realistic, controlled simulations of phishing, smishing, spear-phishing, and baiting attacks with continuous behavioural monitoring and adaptive learning mechanisms. The platform records user interactions— such as link clicks, credential submission attempts, reporting behaviour, and message avoidance—and analyses them to generate transparent vulnerability scores at individual, departmental, and organizational levels. A key contribution of SEAST is its adaptive micro-learning approach, which delivers immediate, contextual training feedback when unsafe user behaviour is detected. This just-in- time learning model reinforces correct security practices more effectively than delayed or generic instruction. Additionally, SEAST incorporates a centralized analytics dashboard that provides real-time insights into campaign performance, user behaviour trends, and risk distribution, enabling data-driven decision-making for administrators. Gamification elements and role-specific simulations further enhance user engagement and long-term retention of security awareness concepts. This paper details the design objectives, system architecture, methodology, comparative evaluation with existing awareness tools, experimental observations, limitations, and future enhancement directions of SEAST. The findings demonstrate that simulation-based, behaviour-focused awareness training significantly improves user vigilance and reduces susceptibility to social engineering attacks. SEAST positions itself as a scalable, cost-effective, and impactful solution for organizations seeking to strengthen cybersecurity defences by addressing the human factor.
🔖 Keywords :️ Social Engineering, Cybersecurity Awareness, Phishing Simulation, Human Firewall, Behavioural Analytics, Adaptive Learning.
Quantum-Safe Encrypted File Sharing Application
👥 Vaibhav, Lavanay, Ritika Sharma
📙 Abstract : Rapid evolution in quantum computing represents a serious danger to traditional forms of public-key cryptography; thus, there is an urgent need to develop quantum restricted (QR) security models. This document provides a comprehensive assessment of the current state of secure file-sharing applications and discusses methods to construct a Quantum-Secure Encrypted File Sharing Application. This design is based on post-Quantum Cryptography using Kyber for Key Encapsulation as a forward-compatible method of achieving secure connections. The review shows that by combining strong Client-Side Encryption with Real-Time Peer-To-Peer (P2P) File Transfers through WebRTC, utilizing the Zero Trust security model, and including Decoy methods such as Honeyfiles, we have created an innovative architecture to address gaps in existing file-sharing solutions. Our intention is to provide impenetrable data integrity and privacy against Classical and Future Quantum Threats, while demonstrating that the architecture is scalable and resilient for exchanging extremely sensitive information.
🔖 Keywords :️ quantum-safe, post-quantum cryptography.
Smart Network Security System: A Review
👥 Anmol Bhalla, Manish Kumar, Neha Tuteja
📙 Abstract : Digital operations require cybersecurity as their fundamental base because cyber-attacks have developed into more sophisticated forms while network systems become more complex and new advanced malware threats keep appearing. The current security threats create problems for Traditional Intrusion Detection Systems (IDS) because signature-based solutions fail to identify the new threat types that keep emerging. Machine Learning (ML) presents a promising alternative which provides adaptable solutions and strong precision rates and detection capabilities for unknown attack types. The research paper examines network security through ML-based approaches while creating a Smart Network Security System which performs real-time intrusion detection. The system operates by using supervised ML models, which include Random Forest, Support Vector Machines, and Gradient Boosting that learn from the UNSW-NB15 dataset to identify network traffic patterns. The paper analyses past research work by identifying datasets for IDS research and evaluates machine learning algorithms through performance metrics and real-time implementation assessment. The research study identifies various areas that require further investigation in its findings. This will help improve network security in real-time systems.
🔖 Keywords :️ Intrusion Detection System (IDS), Machine Learning, Network Security, Real-Time Detection, UNSW-NB15, Ensemble Learning.
Fake Social Media Account Detection Using Machine Learning: A Comprehensive Review
👥 Sagar Bhardwaj, Shivam, Anju Arya
📙 Abstract : Social media platforms have experienced exponential growth enabling billions of users to connect and share content globally. However, this unprecedented expansion has attracted malicious actors who create fake accounts for spam dissemination, impersonation, phishing, and coordinated bot attacks. Traditional rule-based detection systems fail to identify emerging attack patterns and newly crafted fraudulent accounts. Machine learning presents a promising alternative providing adaptive solutions with strong accuracy and detection capabilities for unknown fake account patterns. This research paper examines social media account authenticity through machine learning-based approaches while implementing a Fake Social Media Account Detection System that performs automated classification. The system operates using supervised machine learning models including Logistic Regression, Decision Tree, Random Forest, and Multilayer Perceptron Neural Networks trained on Instagram profile metadata to identify account authenticity patterns. The paper analyzes existing research by identifying datasets and evaluation techniques for fake account detection and evaluates multiple machine learning algorithms through performance metrics and real-world implementation assessment. The study identifies key features predictive of account inauthenticity and demonstrates that neural network approaches achieve 95% accuracy in binary classification tasks. The research identifies various areas requiring further investigation in its findings to improve detection systems in real-world deployment scenarios. This work contributes to developing practical, scalable, and accurate fake account detection mechanisms for social media platforms.
🔖 Keywords :️ Fake Account Detection, Machine Learning, Social Media Security, Binary Classification, Feature Engineering, Neural Networks, Instagram Account Authentication.
SamSec: A Distributed Orchestration Framework for Unified Attack Surface Management and Dynamic Security Testing
👥 Vaishnavi, Janvi Tyagi, Dr.Shakti Arora, Mrs.Shruti Jindal
📙 Abstract : Many organisations with large external asset inventories have one challenge in front of them: there are brilliant sets of sophisticated tools available, but the lack of an efficient combination of these tools remains an issue. Manual data format conversion, tool-to-tool hand-offs, and brittle scripting all contribute to what we call the “Orchestration Gapâ€â€”in the amount of engineering hours that are spent per scan cycle data entry instead of data analysis. This paper introduces the distributed orchestration framework SamSec that overcomes this gap by utilizing a 6-step reconnaissance funnel based on passively discovered subdomains, DNS health analysis, high-speed port scanning, endpoint probing, web crawling, and template-based vulnerability checks. SamSec executes tools in an asynchronously distributed manner via Celery, Redis [10] for this, it also provides normalization of all subsequently executed tools, enabling all of such outputs to be presented in a common JSON schema, it also integrates the MITRE ATT&CK adversarial tactic engine [12] to map the vulnerabilities that are discovered during the process to the adversary's tactics. The progressive funnel architecture brings up to 20X less redundant compute than naïve approaches. SamSec allows you to assess your external attack surface continuously and comprehensively—in minutes—in the same short time frame that it takes engineers to assess external attack surface while avoiding the need to engage in format-conversion tasks. Technology—Attack Surface Management, Tool Orchestration, Automated Vulnerability Assessment, Distributed Systems, DevSecOps, and Asynchronous Processing are example technologies. Technologies—Example technologies include Attack Surface Management, Tool Orchestration, Automated Vulnerability Assessment, Distributed Systems, and DevSecOps, among others, plus Asynchronous Processing.
🔖 Keywords :️ Attack Surface Management, Tool Orchestration, Automated Vulnerability Assessment, Distributed System.