A Hybrid Machine Learning Approach to Real-Time Cyber Threat Detection and Response for Next-Generation Network Security

Abstract

This paper surveys state-of-the-art machine learning (ML) techniques for real-time cybersecurity, focusing on anomaly detection, intrusion detection systems (IDS), Security Information and Event Management (SIEM) integration, and automated response. We review recent literature, summarize strengths and limitations of supervised, unsupervised, and deep learning approaches, and propose a practical, modular ML-driven architecture for real-time threat detection and response that balances latency, accuracy, interpretability, and robustness. We present an experimental plan using public benchmark datasets and discuss evaluation metrics, deployment considerations (scalability, privacy, adversarial robustness), and future research directions.