Boosting False Data Injection Attack Detection with Structural Knowledge

Abstract

State estimation is crucial to the reliable operation of the power grid. Hence, various cyber-physical attacks take advantage of manipulating the state estimation outcome to threaten grid reliability. Such cyber-physical attacks include fuzzing, malware injection and false data injection attack (FDIA). While the traditional residual-based error detection could prevent certain attacks, FDIA is not one of them. This study notices that matrix separation is a powerful tool in terms of FDIA detection. Thus, we cast FDIA detection into the matrix separation framework, embedding two types of structural knowledge. The first one highlights that only some rows in the attack matrix have nonzero values, while the second one emphasizes that the temporal variability of data collected by the same meter is usually small. Our proposed framework yields a structure embedding detection method, and numerical studies highlight its remarkable performance.