Recent Papers

The increasing complexity of cyber threats and the exponential growth of networked systems have made traditional rule-based security mechanisms insufficient for real-time detection and prevention of cyber attacks. This paper explores the application of data science techniquesincluding machine learning, statistical modeling, and big data analyticsfor predicting cyber attacks before they occur. The study analyzes large-scale datasets such as UNSW-NB15 and CICIDS2017 to identify behavioral patterns associated with cyber intrusions. Techniques such as Random Forest, Support Vector Machine (SVM), and Deep Neural Networks (DNN) are evaluated for their predictive accuracy and false alarm rates. Results indicate that ensemble-based models outperform traditional classifiers in identifying potential attacks with an accuracy exceeding 98%. The paper concludes by discussing ethical, legal, and interpretability concerns in deploying AI-based predictive cybersecurity systems. The research conclusively shows that ensemble-based ML models, specifically Random Forest, significantly outperform traditional classifiers, achieving a predictive accuracy of 98.2%. A Deep Neural Network (DNN) also performed strongly with 97.5% accuracy. Key data features like Flow Duration, Destination Port, and Packet Size were identified as powerful predictors of attack probability. While these results present a compelling case for deploying predictive systems in Security Operations Centers (SOCs), the research also highlights critical ethical, legal, and operational challenges. The necessity of data anonymization (per GDPR, India's IT Act), human oversight to manage false positives, and the integration of explainable AI (XAI) are paramount for the responsible and trustworthy implementation of these technologies.

Copyright © 2025 Arjunarao Rajanala, Pagidipalli Sowjanya. This is an open access article distributed under the Creative Commons Attribution License.