Risk-Aware Individual Trajectory Data Publishing With Differential Privacy

Large-scale spatiotemporal data mining has created valuable insights into managing key areas of society and the economy. It has encouraged data owners to release/publish trajectory datasets. However, the ill-informed publication of such valuable datasets may lead to serious privacy implications for individuals. Moreover, as a major goal of data protection, balancing privacy and utility remains a challenging problem due to the diversity of spatiotemporal data. However, the user dimension was not considered for traditional frameworks, which limits the application at the global level as opposed to the user level. Many researchers overcome this issue by assuming that a user in the dataset generates only one trajectory. Actually, a user always generates multiple and repetitive trajectories during observation. Only considering one trajectory for one user may cause insufficient privacy protection at the trajectory level alone, as a user's privacy can be manifested in many trajectories collectively. In addition, it demonstrates strong user correlation when using multiple and repetitive trajectories. If not considered, additional information will be lost, and the utility will be decreased. In this article, we propose a novel privacy-preserved trajectory data publishing method, i.e., IDF-OPT, which can reduce global least-information loss and guarantee strong individual privacy. Comprehensive experiments based on an actual trajectory publishing benchmark demonstrate that the proposed method maintains high practicability in trajectory data mining.