Research On Publishing Trajectory Dataset Via Differential Privacy Based On The Spatio-temporal Correlation

Trajectory data plays an important role in intelligent city and "Internet+" construction,however,it contains a large amount of personal privacy in one trajectory,the privacy protection for the release of trajectory datasets is urgently needed.Differential privacy as a new privacy standards which based on specification definition and solid theoretical foundation,has been gradually applied to different areas of privacy protection.Existing differential privacy method for trajectory dataset release mainly aimed at some of the datasets that have a small location point set,and the location’s time property is not been considered.For the GPS trajectory dataset which carries time property and has a large location point set,the existing method is low efficiency,and added to much noise data to the original dataset.Hence,according to the problem mentioned above,the main goal of this paper is to research on the influence of location’s time property on the accuracy of trajectory dataset’s published results under differential privacy.And we put forward a hybrid release method to spatio-temporal trajectory data via differential privacy(HRST-DP).The main content of this paper is as follows:Firstly,the definition of aggregation degree of spatio-temporal trajectory dataset is proposed to evaluate the sparseness of the dataset,and the aggregation degree’s influence on the accuracy of the published results is also analyzed in detail;Then proposed the concept of generating trajectory set,and given the mathematic description of differential privacy spatio-temporal trajectory dataset publication issue.Secondly,according to the impacts on published results of original spatio-temporal trajectory dataset which has a low aggregation degree,the Spatio-temporal Trajectory Data Aggregation Algorithm(STAA)is proposed.STAA aggregates the similar spatio-temporal location points into a cluster firstly,and then replace original location point in the trajectory with cluster center,so as to improve the aggregation degree of the trajectory dataset.Thirdly,the prefix tree model is used to release the ultimate spatio-temporal trajectory dataset that meet the differential privacy,and to reduce the noise prefix tree’s nodes,the Noise Prefix Tree Build method based on Node Aggregation and Reconstruction(PTBAR)is proposed,PTBAR algorithm includes two parts which are node aggregation algorithm Node-Aggre and the noise added method Merge-Noise.The Node-Aggre algorithm classifies the noise prefix tree nodes in an approximate aggregation order according to the node’s count value.For the clustered tree nodes,the Merge-Noise algorithm is called to combine and reconstruct the adjacent nodes which with similar count values and in the same cluster units.Through the way of multiple iterations,Merge-Noise can increase the probability of nodes merging,and effectively reduce the amount of noise data added to the prefix tree.Finally,we compared the HRST-DP algorithm proposed in this paper with existing methods through the real trajectory dataset,to verify the validity and usability of STAA algorithm and PTBAR algorithm.