An Explainable Clustering-Based Approach for Cyber Situational Awareness on Masquerade Attacks Detection

Abstract

Masquerade attacks pose a significant challenge in cybersecurity, as intruders mimic legitimate user behavior to evade detection. In dynamic, data-intensive environments, traditional intrusion detection systems often struggle to provide both timely and interpretable results, limiting their usefulness for effective Cyber Situational Awareness (CSA). This article presents a clustering-based approach for detecting masquerade attacks using OK-Means—a variant of K-Means optimized for faster convergence—combined with a nearest neighbor classifier and noise reduction techniques. The proposed Intrusion Detection System (IDS) reduces computational overhead while enhancing explainability, leading to more reliable and transparent Cyber Threat Intelligence (CTI) decisions.