Dynamic Pricing for Privacy-Preserving Mobile Crowdsensing: A Reinforcement Learning Approach

Abstract

MCS is an emerging technology that exploits the enormous sensing power of widely used mobile devices to complete sensing tasks in a cost-efficient manner. Among all outstanding issues of current MCS systems, the concern about a lack of privacy protection for the sensing data of participants has drawn increasing attention recently. Various privacy-preserving MCS mechanisms have been proposed for the static scenario where users' privacy protection requirements remain unchanged. In practice, however, users' requirements for privacy protection can be time-varying, which further complicates the design of privacy-preserving MCS. In this article, we first give an overview of multiple promising approaches for privacy-preserving MCS, based on which we make a first attempt to explore privacy-preserving MCS in a dynamic scenario, which is cast as a Markov Decision Process. Specifically, we develop a reinforcement learning based approach, by which the platform can dynamically adapt its pricing policy catering to the varying privacy-preserving levels of participating users. We further use a case study to evaluate the performance of our proposed approach.