Probability Indistinguishable: A Query and Location Correlation Attack Resistance Scheme

As location-based services (LBSs) require users to report their location to obtain services, many people are starting to realize the exposed to high privacy threats. In order to preserve the privacy, a great deal of privacy preserving algorithms is preserved in the last several years. Unfortunately, existing privacy preserving algorithms for LBSs usually mainly consider generalizing or cloaking the location and neglect the correlation probability between the query and location, and the adversary can use the probability to guess the real location. In this paper, based on the concept of differential privacy, we propose a mechanism for achieving probability indistinguishable, and then based on this mechanism a location-shift scheme to obfuscate the correlation between the query and location is proposed. To address the correlation probability obfuscation, we first show the correlation attack model with four potential methods based on the correlation probability. Then we study the proposed attacks on several existing algorithms designed for snapshot as well as continues services and define a formalization of probability indistinguishable to propose a countermeasure with location-shift, which can mitigate this type of attacks. At last, we verify the security of our location-shift scheme with entropy and mutual information, and the empirical evaluations further verify the effectiveness and efficiency of our scheme.