Journal of Cyber Security, Privacy Issues and Challenges

Anti-phishing Frameworks for Safer Digital Ecosystems: A Systematic Review

Kevin Patil — 2026-02-24

The digital ecosystem faces an unprecedented wave of phishing attacks, with over 1.9 million detected incidents in 2024 alone. This systematic review analyzes 30 research papers on phishing detection techniques, spanning classical machine learning to emerging LLM-based systems. A comprehensive taxonomy of detection methods is presented and Random Forest and ensemble methods as current accuracy leaders (up to 99.96%) are identified. Beyond the literature review, empirical validation by implementing two top-performing models Random Forest (Paper 1) and XGBoost (Paper 15) on the largest publicly available dataset (235,795 URLs) is provided.
Key results show that Random Forest achieves 99.9215% accuracy with 77.07/100 adversarial robustness. XGBoost achieves 99.9555% accuracy with a throughput of over 1.5 million samples/sec. Both models surpass or nearly match their original paper baselines on 20× larger data, even after rigorous data leakage removal of 9 suspicious features.

AI-based CCTV Analysis for Student Entry/Exit Monitoring

Akshath S. Karanth — 2026-03-12

Artificial intelligence (AI) has significantly reshaped the role of surveillance systems by enabling them to function beyond simple video recording. In educational institutions, where the safety and tracking of student movement are crucial, the traditional practices of manual attendance or RFID-based access control often fall short in accuracy, speed, and misuse prevention. AI-based CCTV analysis provides an opportunity to automate entry and exit monitoring without requiring physical interaction from students or continuous manual supervision. Deep learning- based models such as YOLO have made real-time person detection and tracking fast, scalable, and highly efficient, even in complex environments with high population movement. This study presents an analysis and evaluation of an AI-based CCTV monitoring system for student entry and exit tracking, along with a review of related research. It highlights the technological backbone including object detection algorithms, Python-based processing, and SQL database integration while critically examining their performance, feasibility, and limitations in real-world institutional settings. In addition to studying the operational benefits, the survey identifies key challenges such as occlusion, lighting variations, privacy concerns, and computational resource demands. The overall objective of this paper is to offer an in-depth understanding of how AI-powered surveillance can improve campus security, automate attendance documentation, and reduce management workload, while outlining opportunities for future enhancement.

Design and Evaluation of a Machine Learning-based Botnet Detection Framework for SCADA Systems in Nigeria

Mission Franklin — 2026-03-19

Supervisory control and data acquisition (SCADA) systems form the backbone of Nigeria’s critical infrastructure, supporting essential services across energy, oil and gas, manufacturing, and refining sectors. Increasing interconnectivity via industrial networks and the Internet exposes these systems to sophisticated cyber threats, particularly botnet attacks. This can disrupt operations, damage data integrity, and have substantial economic and public safety implications. Traditional signature-based security methods are frequently ineffective against developing and zero-day attacks, underlining the necessity for intelligent, adaptive detection solutions. This study presents a machine learning-based botnet detection framework tailored for Nigerian SCADA networks. The framework integrates real-time traffic monitoring, feature engineering, and supervised learning models to identify anomalous and malicious communication patterns. Traffic characteristics such as packet rates, protocol patterns, and flow metrics are identified and examined to improve detection precision and reduce false alarms. Various machine learning algorithms are assessed to evaluate their effectiveness for deployment in real-time, resource-limited SCADA systems. Validation using simulated and real SCADA datasets demonstrates that machine learning models can reliably distinguish normal from malicious traffic, with ensemble and hybrid models showing superior performance. Feature selection further improves computational efficiency without reducing accuracy, supporting practical operational deployment. The study demonstrates the viability of ML-driven botnet detection for strengthening SCADA cybersecurity. It recommends that infrastructure operators adopt adaptive ML-based intrusion detection, continuously retrain models using local traffic, and integrate detection frameworks with national incident response systems. Policymakers should promote standardized SCADA data sharing and capacity-building initiatives to reinforce Nigeria’s overall industrial cybersecurity resilience.

Cloud-based Privacy-Preserving Data Storage

P. Vanitha — 2026-04-08

Cloud computing has significantly transformed the way individuals and organizations store, manage, and access data. Despite its advantages such as scalability, flexibility, and cost efficiency, storing sensitive data in the cloud introduces serious privacy and security challenges. Issues such as unauthorized access, insider threats, cyberattacks, and data leakage have raised concerns about trusting third-party cloud service providers. This paper presents a Cloud-based Privacy Preserving Data Storage (CPPDS) framework designed to ensure confidentiality, integrity, and controlled access to outsourced data. The proposed system integrates encryption mechanisms, secure key management, access control policies, and auditing techniques to protect sensitive information stored in the cloud. Experimental evaluation demonstrates that the framework enhances data privacy while maintaining system efficiency and performance. Furthermore, the framework incorporates role-based authentication to restrict unauthorized access and ensure that only verified users can interact with sensitive data. The system also supports scalability, allowing it to handle increasing volumes of data without compromising security or performance. In addition, real-time monitoring mechanisms are employed to detect and respond to potential security threats promptly. The proposed approach contributes to building a reliable and secure cloud environment suitable for handling critical and sensitive information.

Adaptive Secure Energy-efficient Routing for Emergency MANETs: A Review

Parth Bhatiya — 2026-04-08

Mobile ad hoc networks (MANETs) are widely used in disasters and emergencies where fixed communication infrastructure is unavailable; however, limited battery power, dynamic topology changes, and security threats reduce their efficiency and reliability. This study proposes an adaptive energy-efficient routing strategy for disaster-resilient emergency MANETs. A hybrid routing framework integrating energy-aware routing with dynamic security mechanisms is designed to address these challenges. The system continuously monitors node energy levels and adjusts routing paths accordingly, assigning heavier tasks to high-energy nodes while allowing low-energy nodes to conserve power. Simulations were conducted under varying node densities and mobility conditions to evaluate performance metrics, including energy consumption, network lifetime, packet delivery ratio, and routing overhead. The results show reduced energy consumption and extended network lifetime compared with existing approaches. The proposed method maintains a high packet delivery ratio with moderate routing overhead while ensuring adaptive security. Overall, the strategy effectively balances energy efficiency and secure communication in emergency MANET environments, providing a reliable and practical solution for disaster-resilient wireless networks. Future work may include evaluating performance in larger, heterogeneous networks and integrating advanced, lightweight security techniques and diverse mobility models to enhance performance further.