Explainable AI for Autonomous Threat Detection in Critical Infrastructure Systems

Abstract

Critical infrastructure systems (CIS) including energy, water, transportation, and communication sectors are under continuous cyber-threats from increasingly sophisticated adversaries such as advanced persistent threats (APTs). Artificial intelligence (AI) has enabled autonomous threat detection, yet the opacity of many models poses challenges to trust, transparency, and operational safety. This paper investigates the integration of Explainable Artificial Intelligence (XAI) into AI-driven threat detection for CIS. Using a systematic literature review and thematic analysis of 30 recent studies (2020–2024), the research synthesizes current approaches, frameworks, and implementation challenges. The findings reveal that while XAI improves interpretability, challenges persist regarding trade-offs between accuracy and transparency, adversarial robustness, and human-AI collaboration. A conceptual framework is proposed to guide hybrid XAI adoption, supported by recommendations for human-centric design, policy alignment, and standardization. This work contributes to advancing secure, transparent, and resilient AI systems that reinforce trust and accountability in critical national assets.