This paper proposes a multi threshold attack detection method based on interval estimation methods for networked control systems with false data injection attacks and unknown-but-bounded noises. First, a two-step interval estimation method and the set membership estimation method are used to design the state predictor and estimator for the considered system in order to obtain the corresponding prediction sets and estimation sets. In addition, the zonotopes and ellipsoids are used to describe their outer boundaries. Second, the Monte Carlo method is adopted to calculate the intersection area of the two outer boundaries. Then, based on a large amount of area data obtained from prior experiments, cluster analysis is performed to obtain multiple thresholds for determining different attack scenarios. Finally, the effectiveness and accuracy of the proposed algorithm for detecting false data injection attacks are verified through a simulation example of an automated guided vehicle system model.

This paper proposes a multi threshold attack detection method based on interval estimation methods for networked control systems with false data injection attacks and unknown-but-bounded noises. First, a two-step interval estimation method and the set membership estimation method are used to design the state predictor and estimator for the considered system in order to obtain the corresponding prediction sets and estimation sets. In addition, the zonotopes and ellipsoids are used to describe their outer boundaries. Second, the Monte Carlo method is adopted to calculate the intersection area of the two outer boundaries. Then, based on a large amount of area data obtained from prior experiments, cluster analysis is performed to obtain multiple thresholds for determining different attack scenarios. Finally, the effectiveness and accuracy of the proposed algorithm for detecting false data injection attacks are verified through a simulation example of an automated guided vehicle system model.

This paper investigates a distributed set-membership estimation method which is adaptable to the network topology in the presence of unknown-but-bounded noises. The estimation performance is optimized by adaptively adjusting the neighboring node sets for each sensor. First, a group of set-membership estimators is deployed to obtain individual ellipsoid estimates, and error metrics are introduced to evaluate estimation error for different neighborhood topology configurations. Second, a greedy search algorithm is developed to identify suboptimal neighboring node sets to reduce computational complexity. Finally, an intersection-based strategy is proposed to integrate the estimates from neighboring nodes to obtain the final estimates. Simulation results demonstrate that the proposed algorithm effectively enhances accuracy and robustness while reducing computational complexity.