False Data Injection Attack Detection for Control Systems Based on Correlation Analysis

The correlation of control system data is characterized by its intrinsic dynamics, which is pretty hard to forge by the attacker. In this paper, false data injection detection problem for linear time-invariant systems is studied from a perspective of correlation analysis. Two detection methods are proposed based on targeted data correction construction and analysis. First, a noise encryption-based correlation enhancement mechanism and the optimization-based attack detection method are proposed. Second, a coding-based data correlation construction mechanism is designed and analyzed, and the corresponding detection scheme is proposed. The effectiveness and performance are illustrated by simulation. The proposed correlation-based detection schemes require no control performance sacrifice and can be implemented easily.