A Closer Look: Evaluating Location Privacy Empirically

Abstract

The breach of users' location privacy can be catastrophic. To provide users with privacy protections, numerous location privacy methods have been developed in the last two decades. While several studies surveyed existing location privacy methods, the lack of comparative, empirical evaluations imposes challenges for adopting location privacy by applications and researchers who may not be privacy experts. This study fills the gap by conducting a comparative evaluation among a range of location privacy methods with real-world datasets. To evaluate utility, we consider different types of measures, e.g., distortion and mobility metrics; to evaluate privacy protection, we design two empirical privacy risk measures via inference and re-identification attacks. Furthermore, we study the computational overheads inflicted by location privacy in CPU time and memory requirement. The results are thoroughly examined in our work and show that it is possible to strike a balance between utility and privacy when sharing location data with untrusted servers.