A Risk-Based Cybersecurity Auditing Framework for Smart Grid Infrastructure Using Explainable Artificial Intelligence (XAI)

Abstract

This study suggests a framework for cybersecurity auditing of smart grid infrastructure, which is based on the concept of risk and the use of Explainable Artificial Intelligence (XAI) to produce transparent, prioritized and audit-ready security evidence. The information from public smart grid cybersecurity events was mapped to event labels, asset classes, security-control status, compliance indicators, and cyber-physical impact variables, which were then used to create audit-relevant records. Attack likelihood estimates were made using machine learning models. The attack likelihood, asset criticality, control deficiency score, compliance condition and operational impact were all added together to calculate the final audit risk score. Explainability was used as a technique to identify the most important features that affected each audit decision by applying the SHAP method. The proposed framework achieved 96.38% accuracy, 96.51% precision, 96.38% recall, 96.42% F1-score, and 0.996 ROC-AUC. The results of the ablation showed that the inclusion of the risk component and the XAI component resulted in an improvement in the risk ranking, audit traceability and explanation consistency. The framework translates the cybersecurity detection results into an understandable audit decision, enabling risk-based remediation, compliance review, and an understandable smart grid cybersecurity governance.