XAI-based DoS Attack Detection Framework for Reliable Energy Distribution in Smart Grid Systems

The integration of smart grids has enabled the energy management systems to improve its reliability further and its usefulness alongside resources with an improvement in speed and efficiency level. The growing interconnectivity is, however, paying back with threats from cyberspace, particularly the Denial of Service, which could have devastating repercussions on a global scale with the potential of jeopardizing energy distribution. In this paper, we introduce the xMachine Learning (Explainable-ML) - where Denial of Service (DoS) attacks are monitored in a smart grid context. This framework employs some of the most commonly used machine learning models such as Gradient Boosting Machine (GBM), K-nearest Neighbor (KNN), Naive Bayes (NB), and Random Forest (RF) towards comprehensively tracking and categorizing as legitimate or illegitimate requests the volume of traffic on its network. The maximum accuracy of the GBM algorithm stands at 0.9021, thereby assisting in mitigating or, better still, responding to attack threats in the energy distribution network. We have also employed Explainable-AI (XAI) techniques of SHAP and LIME in our models to provide an element of interpretability and absence of ambiguity in the decisions made by the models. This is a trust booster among grid operators, as these models readily explain performance metrics like accuracy, precision, recall, Fl-score, and ROC curves, thus supporting our framework as a plausible approach towards enhancing security and Intearity,