PROJECT TITLE :

Eclipse: Preserving Differential Location Privacy Against Long-Term Observation Attacks

ABSTRACT:

Location privacy can be achieved through the use of mechanisms that are built on geo-indistinguishability. These mechanisms allow users to provide obfuscated location data to providers of location-based services while still being able to use the services as intended. On the other hand, these mechanisms are susceptible to attacks based on inferences. If an attacker has prior knowledge of a user's obfuscated locations, in particular, then they are able to infer the actual locations of the user by carrying out long-term observation attacks. In the field of differential location privacy, unfortunately, the question of how to defend against long-term observation attacks remains unanswered. In this paper, we begin by demonstrating the dangers posed by long-term observation attacks to the existing mechanisms. In light of these vulnerabilities, we come up with a novel mechanism that we call Eclipse. This mechanism closes the gap between the protection of locations and the usability of services. To be more specific, we obfuscate locations by using geo-indistinguishability and k-anonymity, and we hide each location based on an anonymity set. As a consequence of this, our mechanism is able to effectively perturb the distribution of locations and prevent leakage even when subjected to long-term observation attacks. In addition, the set of possible outputs is used in order to reduce the negative effects on the usability and correctness of the system. Utilizing differential privacy allows us to formally define, and then rigorously prove, the safety of the proposed mechanism. In addition, we put the proposed mechanism into action and run a number of experiments on datasets taken from the real world in order to demonstrate the mechanism's usefulness and effectiveness.