Cyber Attack Simulation and Detection in Digital Substation

Abstract

Transforming electrical grids into digital systems has brought many advantages, but it has also introduced new vulnerabilities that cyber attackers can exploit. Therefore, early detection of these attacks is crucial to minimize the impact on power grid operations. This paper presents the results of our investigation into the simulation and detection of cyber attacks in digital substations. Our study focuses on comparing multiple machine learning algorithms for detecting replay attacks and false data injections. The results of our study show that the best model for replay attack detection is the Logistic Regression with an accuracy of 94%. On the other hand, for false data injection detection, multiple models show high precision, recall, F1-score, and accuracy, with the best model in terms of computation time being Support Vector Machine. Our findings provide valuable insights into using machine learning algorithms to simulate and detect cyber attacks in digital substations.