A Trajectory Data Privacy Protection Algorithm Based On Trajectory Segmentation

With the popularity of location-based service applications,application providers have accumulated a large amount of user's trajectory data.Through data analysis,researchers have extracted a lot of useful information from the published trajectory data set,which has broad application prospects in traffic monitoring,urban planning,mobility management and other fields.However,direct release of trajectory data with rich private information can lead to serious privacy leakage issues,so it is necessary to process the trajectory data before release.Due to the large-scale,high-dimensional,and rich background knowledge of trajectory data,the research on privacy preserving technology in trajectory data publishing is facing severe challenges.This thesis focuses on the privacy preserving technology in trajectory data publishing,and a trajectory segment-based trajectory data privacy preserving algorithm is proposed,and the algorithm includes two sub-algorithms.Firstly,aiming at the problem that the traditional method discards too many spatiotemporal points and the number of trajectories in the equivalence class may be too small,an equivalence class partitioning sub-algorithm based on trajectory segment filling is proposed,which divides original trajectory data set into several equivalence classes.If the number of trajectories that original equivalence class contains is fewer than the threshold,some trajectory segments will be filled into the original equivalence class: select several equivalence classes as the divided equivalence classes(these equivalence classes' time intervals are superset of the time interval of the current equivalence class),and then the trajectory segments sliced from the divided equivalence classes will be filled into the current equivalence class.Secondly,aiming at the problem that the traditional method has too large spatiotemporal disturbance distance and too many deleted trajectories,a cluster group construction sub-algorithm based on trajectory segment clustering is proposed.Through this algorithm,every equivalence class will be transformed into several cluster groups.First of all,the trajectories included in the equivalence class are divided into several candidate cluster groups according to the position of the trajectory at the start time of the cluster group.Then,traverse every candidate cluster groups,and the trajectories that that make up current cluster group and the end time of current cluster group are determined.Finally,the trajectories that are not added to the cluster group will be split into multiple trajectory segments,and these segmentswill be filled into the corresponding equivalence classes.In the experimental part,the proposed algorithm(composed of the equivalence class partitioning algorithm and the cluster group construction algorithm)is compared with the classical NWA algorithm from the perspectives of security,availability and execution efficiency.Experimental results show that the proposed algorithm has lower risk of privacy leakage,smaller data loss,and less running time when the data set contains a large number of trajectories and the time intervals of the trajectories are not much different.