The privacy preserving method for dynamic trajectory releasing based on adaptive clustering

The traditional trajectory privacy preserving methods are static releasing methods, only focusing on the trajectories in a specific time span. Therefore, the traditional methods need to recalculate the last released results, in dealing with dynamic trajectory releasing issues as the new time slice arriving. That would increase the computing costs of dynamic trajectory releasing. Furthermore, the static methods treat the trajectory as the basic unit of clustering, and that is a coarse granularity unit. For there exist the different parts among the trajectories, the formed k equivalence classes would have a large trajectory generalization area, severely reducing the availability of the released trajectories. Thus, the dynamic trajectory releasing method based on the adaptive clustering is proposed. This method utilizes the designed Gibbs sampling-cluster method to detect the RR (Representative Region). By which the issue of unaligned sampling time caused by the different moving speed and the sampling frequency can be resolved. And the detected RRs are treated as the critical region to be protected. In this paper, we utilize the proposed fitness function, which could evaluate the rationality of the equivalence classes, to cluster the RRs into GR (Generalization Region), and guarantee the trajectory segments between GRs k-anonymity. The adaptive adjustment policy is adopted in the RR and GR detection process. The adjustment process of RR and GR could effectively reduce the computing cost in trajectory releasing, by considering the relevance between the successive time slices. By the experimental comparison, the effectiveness and rationality of the proposed method and is verified.