Research On Differential Privacy Protection Algorithm For Traffic Trajectory Data Release

The existing location-based service system has collected a large number of user trajectory data.If these data are released directly by data manager without any processing,the user’s privacy will be disclosed.Differential privacy technology can protect data privacy,but how to apply it to location-based service system is also a challenge.According to the characteristics of its data,such as large scale,high dimension and sparse,this paper focuses on how to realize the protection of traffic trajectory data and mining frequent trajectory sequence on the premise of satisfying differential privacy protection.The specific research work is as follows:(1)This paper proposes a method of traffic trajectory data release based on differential privacy protection.Firstly,a noiseless prefix tree is constructed based on the spatiotemporal trajectory data.The prefix tree stores the common prefix information of the track,including timestamp,location and count.Then,according to the count of nodes in the tree to save the prefix subsequence,using incremental privacy budget allocation mechanism,Laplace noise is added to the count of nodes in the prefix.According to the shortest reachable time matrix between location,designed spatiotemporal dimension reduction method reduce the dimension of location and time combination and avoid dimension explosion without consuming any privacy budget.and threshold function are set for each layer of nodes.Through pruning,nodes are reasonably reserved to ensure the consistency constraints between the count of parent nodes and child nodes in the tree structure,so as to build a noise prefix tree that meets the constraints.Finally,the trajectory dataset is synthesized from the noise prefix tree.The proposed algorithm is validated on four different scale trajectory datasets.The data is the real datasets provided by Shenzhen Metro,which contains the card swiping data of 2.8 million passengers in 121 metro stations within 24 hours.Analyze the algorithm from three aspects,a large number of experimental analysis shows that compared with the previous research work,the trajectory data results published by the model algorithm proposed in this paper have better practicability,and the algorithm has higher efficiency and scalability.(2)A frequent sequential pattern mining algorithm for differential privacy protection is proposed.In order to reduce the noise,this paper stores the trajectory in the prefix tree before mining the frequent trajectory sequence,adds noise,prunes and other operations tothe tree node count to protect the privacy of the trajectory data,and finally digs the frequent non-continuous trajectory sequence through the frequent sequence mining algorithm.The number of true positive and utility loss rate are used to verify the effect of frequent sequence mining.In the real traffic trajectory dataset,a large number of experiments show that the purified location trajectory dataset output by the algorithm has high availability,and the frequency sequence obtained by mining has high quality.This paper focuses on differential privacy protection and frequent sequence mining of traffic trajectory data,and evaluates,verifies and analyzes the feasibility and efficiency of the proposed algorithm from various aspects.It makes up for the shortcomings of the existing research methods and has certain research value.