An Explainable CNN-based Intrusion Detection System for Enhanced Smart Grid Security

The evolution from traditional power grids to smart grids, driven by the wide adoption and rapid integration of advanced sensing and communication technologies, introduces new business opportunities alongside critical technical challenges, particularly in the realm of cyber resilience in disaster situations. Coping with the increasing spectrum of cyber threats and their sophisticated evasion techniques is among the most critical challenges due to the devastating impact on individuals and the society in case of a successful attack. Therefore, we propose in this paper a robust intrusion detection system (IDS) specifically designed for smart grid environment and constraints. This IDS employs convolutional neural network (CNN) method to effectively identify and neutralise potential security threats, and the obtained evaluation results are promising. Moreover, our CNN model is complemented by incorporating the SHapley Additive exPlanations (SHAP) algorithm to improve the transparency of the decision-making process.