On Privacy of Dynamical Systems: An Optimal Probabilistic Mapping Approach

We address the problem of maximizing privacy stochastic dynamical systems whose state information is released through quantized sensor data. In particular, we consider setting where about system obtained using noisy measurements. This data and transmitted to a (possibly untrustworthy) remote station public/unsecured communication network. aim at keeping (part of) private; however, because network (and/or station) might be unsecure, adversaries have access data, which can used estimate state. To prevent such from obtaining an accurate estimate, before transmission, randomize additive random vectors, send corrupted instead. design joint probability distribution these vectors (over time window) minimize mutual (our metric) between some linear function (a desired private output) randomized for level distortion-how different measurements distorted are allowed be. pose synthesising as convex program subject constraints. Simulation experiments presented illustrate our scheme.