A Novel Personalized Differential Privacy Mechanism for Trajectory Data Publication

With the development of smart city, more organizations analyze people's trajectory data, so as to provide better location-based services. However, publishing the original trajectory data directly raises serious privacy threats to individuals. As a kind of powerful framework for providing formal and strong privacy guarantees, differential privacy has been applied in the trajectory data publication. Nevertheless, the existing approaches assume that individuals require the same privacy preference, and thus the same level of privacy protection is provided for all individuals, which leads to insufficient privacy guarantee is provided for some individuals, while the other individuals received excess privacy protection. This paper assumes that individuals require different level of privacy and propose a personalized differential privacy publication mechanism for trajectory data. We apply the Hilbert curve to extract the distribution characteristics of the trajectory data at each time and propose a personalized different privacy generalization algorithm for trajectories with different privacy preferences. Through extensive experiments on real world trajectory dataset, we show that this mechanism provides better tradeoff between data privacy and utility compared with the uniform differential privacy based methods.