Adaptive Machine Learning-Driven Cybersecurity: Enhancing Real-Time Threat Detection and Response
A. Jeyaram, A. Muthukumaravel · 2024
Traditional methods of sensitive data protection are static and reactive, which makes them often inadequate in the dynamic and always-shifting cyber threat environment of today. The paper proposes a dynamic and adaptable approach to combine machine learning (ML) at the intersection of cybersecurity and data engineering to address the limitations of conventional solutions. Using ML algorithms to continuously learn from data patterns, the proposed system ensures proactive security mechanisms and allows real-time adaptation to new threats. The proposed system uses supervised and unsupervised learning techniques to effectively distinguish between regular and suspicious activities, hence enhancing cyber resilience. Furthermore, the self-sufficient response to threats and mitigation measures of the proposed system streamlines incident response procedures, which results in faster reaction times and reduced false positive rates. Comparative performance metric results, which indicate reaction times of 5 seconds for low-severity threats, 10 seconds for medium-severity threats, and 15 seconds for high-severity threats, highlight the improved efficacy of the proposed system. At a false positive rate of 5% and a detection accuracy of 95%, the proposed system surpasses existing systems and shows its potential to significantly increase cybersecurity defences in the dynamic threat landscape of today.