Detecting Cyber Attacks in a Cyber-physical Power System: A Machine Learning Based Approach

Abstract

As a result of digitalization and the addition of intelligent devices, the conventional power system has been reformed into a cyberphysical power system (CPPS) with a close interlink between information and energy flow. The CPPS uses pervasive sensing technology, sophisticated measurement technology, and robust information processing technology to achieve the observability and controllability of the electric grid. However, CPPS is more prone to attack than any prior single-structured system due to the high volume of smart device accesses and frequent exchange of information. With the help of the cyber subsystem, malware and hackers can target the CPPS, which can then be fatal to the physical system that supplies energy. In view of this, the paper focuses on detecting and identifying the cyber attacks on the CPPS. The paper proposes an intrusion detection system (IDS) employing a decision tree for classifying different types of cyber attacks launched on the CPPS. The evaluation metrics such as accuracy, precision, recall, and F1 score are computed for different types of cyber attacks to show the effectiveness of the proposed IDS. Finally, from the results, it can be claimed that the proposed IDS is successful in identifying the various cyber attacks on CPPS in different test scenarios.