Optimal sporadic location privacy preserving systems in presence of bandwidth constraints.

ABSTRACTVarious Location Privacy-Preserving Mechanisms (LPPMs) have been proposed in the literature to address the privacy risks derived from the exposure of user locations through the use of Location Based Services (LBSs). LPPMs obfuscate the locations disclosed to the LBS provider using a variety of strategies, which come at a cost either in terms of quality of service, or of resource consumption, or both. Shokri et al. propose an LPPM design framework that outputs optimal LPPM parameters considering a strategic adversary that knows the algorithm implemented by the LPPM, and has prior knowledge on the users' mobility profiles [23]. The framework allows users to set a constraint on the tolerable loss quality of service due to perturbations in the locations exposed by the LPPM. We observe that this constraint does not capture the fact that some LPPMs rely on techniques that augment the level of privacy by increasing resource consumption. In this work we extend Shokri et al.'s framework to account for constraints on bandwidth consumption. This allows us to evaluate and compare LPPMs that generate dummies queries or that decrease the precision of the disclosed locations. We study the trilateral trade-off between privacy, quality of service, and bandwidth, using real mobility data. Our results show that dummy-based LPPMs offer the best protection for a given combination of quality and bandwidth constraints, and that, as soon as communication overhead is permitted, both dummy-based and precision-based LPPMs outperform LPPMs that only perturb the exposed locations. We also observe that the maximum value of privacy a user can enjoy can be reached by either sufficiently relaxing the quality loss or the bandwidth constraints, or by choosing an adequate combination of both constraints. Our results contribute to a better understanding of the effectiveness of location privacy protection strategies, and to the design of LPPMs with constrained resource consumption.