Distinguishing between cyber injection and faults using machine learning algorithms

Concern about cyber-security is growing worldwide with the advancement of smart control and networking systems in the cyber-physical layer of power systems. Detection of the stealthy False Data Injection Attack (FDIA) is getting more complicated when the system's behavior during external disturbances, e.g. faults are considered. In this paper, a machine learning algorithm based approach is proposed to detect and distinguish between stealthy FDIA in the state estimator and faults in power systems. The detection rate and false positive rate obtained by using different state-of-the-art classifiers show that the proposed approach can successfully distinguish between cyber injections and faults in power systems. Cyber injection and faults are introduced in a state estimator model simulated in MATLAB and an open source machine learning tool, WEKA is utilized to distinguish the injection and faults from the developed dataset.