Intrusion Detection Techniques for Cyber-physical Systems
Comfort Oluwaseyi Folorunso, Lateef Adesola Akinyemi, Wasiu Adewale Raheem, Olumide Francis Odeyinka, Eniola Sodiq Olaleye, Oluwafemi Ipinnimo · River Publishers eBooks · 2025
The dynamic nature of cyber security threats necessitates innovative approaches to safeguard networks and systems. This work examines intrusion detection techniques to prevent attacks that can harm cyber-physical systems (CPS). Intrusion detection techniques for CPS employ cutting-edge detection meth ods such as behavior analysis and anomaly detection algorithms to identify and prevent infiltration attempts. This work emphasizes achieving optimiza tion, digital operation, and safety in a CPS using a power generation plant case study. Nine different machine learning algorithms, including CatBoost, deci sion tree, XGBoost, lightGBM, random forest, AdaBoost, XGBoost, logistic regression, k-nearest neighbors (KNN), and naive Bayes were employed in this study and their performance on attack detection have been examined. The HIL-based augmented ICS (HAI) security dataset is used. The random 436 forest gave the best results with AUC, sensitivity, accuracy, precision, and F1 of over 99.31%, followed by the CatBoost (over 99.1%) and the XGBM (over 98.86%). The lessons learned and future directions are presented.