Explainable SCADA-Edge Network Intrusion Detection System: Tree-LIME Approach
Cosmas Ifeanyi Nwakanma, Love Allen Chijioke Ahakonye, Taesoo Jun, Jae Min Lee, Dong‐Seong Kim
Abstract
This study presents an Explainable AI (XAI) conceptual framework for intrusion detection in the SCADAedge network based on an optimizable tree algorithm for attack classification and a Decision Tree-based local interpretable model agnostic explanations (LIME) approach. This approach creates plausible explanations that capture nonlinear interactions between data features, resulting in more reliable interpretability, fidelity, and efficiency, as demonstrated in experimentation with the Edge dataset. Experimentation shows the Tree-LIME XAI approach giving faithful and local explanations of the 12 relevant features of the recently publicly available Edge-IIoTset.