Differentially Private Fingerprinting for Location Trajectories

Location-based services have brought significant convenience to people in their daily lives. Services like navigation, food delivery, and carpooling frequently ask for location data from users. On the other side, researchers and businesses are eager to acquire those data (that is collected by location-based service providers) for various purposes. However, directly releasing those data causes privacy concerns since location data contain users' sensitive information, e.g., regular moving patterns and favorite spots. To solve this, we propose a system that protects users' location data under differential privacy and prevents unauthorized redistribution at the same time. Observing high amount of noise introduced to achieve differential privacy, we implement a novel post-processing scheme to regain data utility. In addition, we also propose a novel fingerprinting scheme as a part of the post-processing (to detect unauthorized redistribution of data). Our proposed fingerprinting scheme considers correlations in location datasets and collusions among multiple parties, which makes it hard for the attackers to infer the fingerprinting codes and avoid accusation. Using the experiments on a real-life location dataset, we show that our system achieves high fingerprint robustness against state-of-the-art attacks. We also show the integrated fingerprinting scheme increases data utility for differentially private datasets, which is beneficial for data analyzers in data mining.