Leveraging unique CPS properties to design better privacy-enhancing algorithms

Abstract

Cyber-Physical Systems (CPS) have unique properties that can be exploited to design new privacy-enhancing technologies that minimize the negative impact to the utility of CPS. In this paper we show two examples of these properties. The first example looks at how differential privacy degrades CPS performance due to the large noise addition, and we then show how the inherent noise of CPS can be leveraged to reduce the additional noise added by differential privacy algorithms, and therefore, minimize the negative impact on the system utility and safety. In the second example we look at the ability to sample at sensor readings on demand, and how this fiexibility can be used to design adaptive sensor sampling algorithms that hide sensitive information without the need to add noise.