Affiliations:
Department of Computer Science, Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah, Saudi Arabia
A large volume of sensitive raw data is continuously collected from data acquisition systems within the monitoring and control networks of the power grid to support key applications in the centralized control system for smart grid operation, management, and planning. Although these communication networks provide wide-area and high-speed connectivity, they also increase the risk of cyberattacks that threaten the grid’s critical physical infrastructure. The current centralized approach to intrusion detection cannot meet the strict quality-of-service requirements of latency-sensitive applications, and the growing size and complexity of learning models further increase communication and computation demands. Edge-intelligent access points offer a promising solution by enabling lightweight learning models to run close to data sources and provide fast responses to protect the core infrastructure. This paper proposes a lightweight machine learning-based intrusion detection system to support a shift toward distributed learning. Six learning models are used for feature extraction and classification, and the Synthetic Minority Oversampling Technique (SMOTE) is applied to balance the dataset. The model’s performance is evaluated under binary and multiclass classification scenarios, and the results show excellent accuracy, short training time, and strong ability to distinguish various attack types, demonstrating its suitability for smart grid environments.
Smart grid security, Intrusion detection, Lightweight models, Edge intelligence, Machine learning
https://doi.org/10.21833/ijaas.2026.03.006
Nassef, L. (2026). A lightweight machine learning-based intrusion detection system for smart grids. International Journal of Advanced and Applied Sciences, 13(3), 52–67. https://doi.org/10.21833/ijaas.2026.03.006