Gaussian Mechanism Design for Prescribed Privacy Sets in Data Releasing Systems

Abstract: The data transmitted by cyber-physical systems can be intercepted and exploited by malicious individuals to infer privacy-sensitive information regarding the physical system. This motivates us to study the problem of preserving privacy in data releasing of linear dynamical system using stochastic perturbation. In this study, the privacy sensitive quantity is the initial state value of the system. For protecting its privacy, we directly design the covariance matrix of a Gaussian output noise to achieve a prescribed uncertainty set in the form of hyper-ellipsoids. This is done by correlated noise and through a convex optimization problem by considering the utility of released signals. Compared to other available methods, our proposed technique for designing the Gaussian output noise provides enhanced flexibility for system designers. As a case study, the results are applied to a heating ventilation and air conditioning system.