Explainable Artificial Intelligence for Blockchain-Based Intrusion Detection Systems in Critical Infrastructure

Abstract

The increasing reliance on Artificial Intelligence to secure critical infrastructure necessitates intrusion detection systems that need to be accurate, transparent and also tamper-resistant. This article proposes a new framework that integrates Explainable Artificial Intelligence (XAI) with blockchain to enhance the trustworthiness, explainability, and integrity of Intrusion Detection Systems (IDS) in industrial environments. The system employs Random Forest classifiers in combination with SHAP and LIME to provide human-understandable explanations of the detected anomalies. For safe and immutable alert logging, the system employs smart contracts within a permissioned blockchain network (e.g., PBFT or Raft). Experimental comparison on the TON_IoT Modbus dataset, a simulation of actual industrial telemetering, suggests that the new system achieves 97.31% detection accuracy with balanced precision and recall. Feature importance analysis provides important contributors to classification outcomes, and blockchain-based logging provides tamper-proof record-keeping with acceptable latency. The above architecture provides enhanced detection transparency, and forensic auditability also provides an extensible basis for a reliable, explainable IDS deployment across critical infrastructure sectors such as smart grids, healthcare, and industrial control systems.