Towards Early Cyberattack Discrimination in the Smart Grid Using Statistical Approach with Machine Learning Techniques

Abstract

Smart Grid has been exposed to cyberattacks that penetrate the Supervisory Control and Data Acquisition systems, causing privacy and access control violations on security infrastructures. It is critical to address this issue with an early attack detection mechanism with a reduced missing rate and less response time that discriminates cyberattacks against natural and man-made events. Developing a robust intrusion detection system is still a challenging research problem to prevent the smart grid system from sabotage and processing downtime. This paper aims to develop a robust Anomaly-Based Intrusion Detection System using the statistical approach with a machine learning classifier to discriminate cyberattacks from natural faults and man-made events to avoid blackouts and cascading failures. The novel mechanism of Statistical Approach with a Machine Learning classifier (SAML) based on Neighborhood Component Analysis, ExtraTrees, and AdaBoost for feature extraction, bagging, and boosting respectively is proposed, to optimally tune the hyperparameters for early discrimination of Triple class events. The proposed model is tested on the publicly available Industrial Control Systems Cyber Attack Power System Triple Class Dataset. The test results provide a higher detection accuracy with a low missing rate, false alarm rate, and less response time than the existing approaches.