Explainable Artificial Intelligence for Intrusion Detection in Cyber-physical Systems
Yakubu Tsado, Shokooh Khandan, Segun Isaiah Popoola · River Publishers eBooks · 2025
Cyber-physical systems (CPSs) play a critical role in modern infrastructure and are regarded as a leading technology in advancing Industry 5.0 – a paradigm that integrates human interaction within CPS to enhance human capabilities, addressing concerns from Industry 4.0 where advanced technologies could substitute human roles. Due to the massive data generated by CPS, various predictive machine learning (ML) and deep learning (DL) models have been deployed for performance monitoring, optimization, preventive maintenance, and threat detection. However, a significant challenge in adopting these models in critical domains like CPS is their “black box” nature, which makes them difficult to interpret and trust. This chapter explores how explainable AI (XAI) can be utilized to enhance the interpretability of intrusion detection systems (IDS) within CPS. Firstly, we review recent studies on XAI methods applied to attack detection and provide a taxonomy of XAI techniques, encompassing model-specific and model-agnostic approaches that can be applied to ML and DL models. This taxonomy supports transparent insights into each model’s learning and decision-making processes. 78 Furthermore, we discuss local and global explainability, where local explainability focuses on how individual features influence specific predictions, and global explainability reveals how features affect the overall model behavior. These insights aim to provide clearer, actionable explanations to strengthen IDS deployment within CPS.