A Coding-based Differentiation Approach for False Data Injection Attacks and Sensor Faults in Networked Control Systems

Based on the coding scheme, this paper proposes a method to distinguish false data injection (FDI) attacks and sensor faults in networked control systems. Since attacks and faults occur on physical devices and cyber layers respectively, the coding scheme is capable of amplifying attack-induced residuals, while keeping fault-induced residuals unchanged. We utilize the false positive and negative rates to evaluate the effectiveness of the coding scheme in achieving threat differentiation. Accordingly, we design a suitable threshold for the residue-based detector, and derive a sufficient condition for the coding matrix, such that the above two indicators can be guaranteed to be sufficiently low. Furthermore, from the perspective of mathematical statistics, we analyze the possibility of an intelligent attacker cracking the coding scheme by eavesdropping the transmission data. In this case, the time-varying coding matrix should be adopted to ensure its effectiveness in distinguishing attacks and faults. Finally, some numerical simulations are given to verify the theoretical results.