Differentially Private Interval Observer Design with Bounded Input Perturbation

Abstract

Real-time data processing for emerging systems such as intelligent transportation systems requires estimating variables based on privacy-sensitive data gathered from individuals, e.g., their location traces. In this paper, we present a privacy-preserving interval observer architecture for a multiagent system, where a bounded privacy-preserving noise is added to each participant’s data and is subsequently taken into account by the observer. The estimates published by the observer guarantee differential privacy for the agents’ data, which means that their statistical distribution is not too sensitive to certain variations in any single agent’s signal. A numerical simulation illustrates the behavior of the proposed architecture.