Interpretable Machine Learning Models for Cyber Threat Detection

作者：Uzair Aslam Bhatti, Yonis Gulzar, Yazeed Yasin Ghadi, Vаlisher Sаpаyev, Tursunov Muhiddin, Akmaljon Khakimov Mirzaganiyevich, Hajiyev Ulugbek Kuzibayevich, Gularam Masharipova · 发表于：Advances in computational intelligence and robotics book series · 年份：2026 · DOI：10.4018/979-8-2600-3465-1.ch004 · 研究领域：Explainable Artificial Intelligence (XAI)、Adversarial Robustness in Machine Learning、Imbalanced Data Classification Techniques

The growing volume, diversity and sophistication of cyber threats rendered traditional signature-based and rule-driven security mechanisms incapable of protecting modern digital infrastructures. To address these complexities, machine learning has been regarded as an effective paradigm for detecting cyber threats due to its capability to process a large amount of heterogeneous security data and detect both previously known and new attack patterns. But many high-performing machine learning (ML) models act as black boxes, causing transparency, trustworthiness, and practicality issues in security-susceptible environments. This challenge has stimulated growing interest in interpretable machine learning, which aims to provide human analysts with model predictions that are understandable, explainable and actionable.