Intelligent Threat Detection Systems: A Survey of Machine Learning Approaches in Cybersecurity

Abstract

The escalating sophistication of cyber threats has rendered traditional signature-based detection systems increasingly inadequate in protecting modern digital infrastructures. Machine learning (ML) and artificial intelligence (AI) have emerged as powerful enablers of intelligent threat detection, capable of identifying novel attack patterns through behavioral analysis and anomaly detection. This paper surveys ML-based approaches applied to cybersecurity threat detection, examining methodologies for intrusion detection, malware classification, phishing identification, and advanced persistent threat (APT) detection. We analyze the effectiveness of supervised classifiers, unsupervised anomaly detection algorithms, and deep learning models in processing network traffic, system logs, and endpoint telemetry data. The survey further addresses key challenges including adversarial machine learning, class imbalance, real-time processing requirements, and the interpretability of detection models in operational security contexts. Our findings indicate that ensemble methods and hybrid architectures combining signature-based and behavioral approaches yield superior detection performance compared to standalone models