Trajectory Synthesis Method Based on Personalized Local Differential Privacy

Trajectory datasets are widely used in location-based services, providing convenience to users and posing privacy risks. The traditional approach to trajectory dataset protection relies on the processing of a central server and assumes the credibility of the server, which may be too strict. The proposal of local differential privacy overcomes this challenge, only the user can access the original data. This paper introduces a new trajectory synthesis method based on personalized local differential privacy, TSPLDP, to publish synthetic trajectory datasets assuming that the central server is not fully trusted. On the client side, trajectory points are encoded based on the quadtree index and perturbed with personalized mechanism. On the server side, the trajectory points distribution is estimated, and then the trajectory points are generalized to synthesize a new trajectory dataset for publication. The actual trajectory dataset is used to evaluate our method. The experiment results show that TSPLDP can effectively protect the privacy of trajectory data, and the synthesized trajectory dataset has better utility.