Comprehensive Analysis of SCADA System Data for Intrusion Detection Using Machine Learning

Abstract

This report investigates the implementation of advanced machine learning models within Supervisory Control and Data Acquisition (SCADA) systems to enhance intrusion detection capabilities and system security. By utilizing models such as CatBoost and XGBRegressor, which excel in processing complex, non-linear data, the study demonstrates significant improvements in predicting and managing operational states in wind turbines. The incorporation of Explainable AI (XAI) techniques, particularly SHAP values, further provides transparency in model decisions, fostering trust among stakeholders. Recommendations are provided for effective model integration, deployment with XAI features, and necessary policy enhancements to ensure the secure, reliable, and ethical use of AI in critical infrastructure environments.