Privacy-Preserving State Estimation with Application to Target Tracking

This paper studies the design of a privacy-preserving optimal state estimator for discrete-time linear systems.Insome traditional methods, such as noise injection, privacy is protected by adding noise to observations and the resulting data is deceptive information. The features of the proposed privacy protection in this paper are twofold. (i) Privacy is protected without providing deceptive information, that is, the information of the resulting protected observations is authentic. The privacy protection consists of two steps. First, the direction deviation of the observations, rather than the raw observation, is computed. Then, this deviation is random and is not always transmitted to the estimator. (ii) An optimal estimator is designed with desired privacy-preserving degree. By tuning a privacy-protection parameter, a given privacy-preserving degree and an estimation accuracy upper bound can be achieved simultaneously. Finally, drone-tracking experiments are provided to demonstrate the effectiveness of the proposed method, and some comparisons with the existing methods are presented.