False Data Injection Attacks Against Error-State Kalman Filter-Based IMU-GPS Fusion Models

This study investigates the vulnerability of error-state Kalman filter (ESKF) based IMU-GPS fusion localization systems to false data injection attacks. Through systematic analysis of the ESKF's internal structure and update mechanism, we derive the recursive relationships governing position and velocity deviations resulting from injected false data. Our findings reveal that position deviation is a function of previous deviation, lateral velocity, and injected data, while lateral velocity is influenced by the differential between injected data and current deviation, as well as the previous lateral velocity. Leveraging these insights, we propose a novel false data injection strategy formulated as a constrained optimization problem, designed to minimize attack energy while achieving desired deviations. Building upon the derived model, we systematically identify optimal solutions that achieve desired position deviations with minimal data injection. Extensive simulations conducted under diverse scenarios demonstrate the efficacy of our proposed method. Furthermore, our results elucidate discrepancies between theoretical predictions and experimental outcomes, highlighting the complexities inherent in real-world applications. This research underscores the need for more secure multi-sensor fusion localization algorithms.