Feature Selection–Based Detection of Covert Cyber Deception Assaults in Smart Grid Communications Networks Using Machine Learning

Abstract

The integration of computing and modern wireless communications techniques is enabling prolific intelligent monitoring and efficient control of electric power systems in the frameworks of smart grids. In parallel, an enhanced reliance on such technologies has increased the susceptibility of today’s smart grids to cyber-assaults. Recently, a new type of assault, termed <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">covert cyber deception assault</i> , has been introduced to infringe upon the integrity of smart grid data. Such assaults are designed and initiated by hackers who have considerably good knowledge of the power network topology and the security measures in place, and therefore, these assaults cannot be effectively detected by the bad-data detectors in traditional state estimators. In this paper, we propose a supervised machine learning–based scheme to detect a covert cyber deception assault in the state estimation–measurement feature data that are collected through a smart-grid communications network. The distinctive characteristic of the paper is that we use a genetic algorithm–based feature selection in our scheme to improve detection accuracy and reduce computational complexity. The proposed detection scheme is evaluated using standard IEEE 14-bus, 39-bus, 57-bus, and 118-bus test systems. Through performance analysis, it is shown that the proposed scheme provides a significant improvement in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">covert cyber deception assault</i> detection accuracy, compared with existing machine learning–based schemes.