Estimation Against Adversarial Attacks for Discrete-time Systems with Intermittent Observations

In this paper, we address the secure estimation problem against false data injections for a class of nonlinear discrete-time networked control systems, where the observation travels along a lossy communication network. Through the use of intermediate variable, an estimation scheme, capable of reconstructing state and attacking signals simultaneously, is first derived to tackle false data attacks on the system's dynamics in terms of linear matrix inequalities. This condition is further expanded to the case where an extra malicious adversary has access to the network and undermines the feedback system by injecting false data into observation packets. It is shown that the resulting estimators achieve uniformly ultimately bounded estimation errors for both state and attacking signals. A numerical example is also included to show the effectiveness of the current results.