Research On Trajectory Privacy-Preserving Theory And Key Technology

Intelligent network connected vehicle trajectory data is of great value for traffic management,business location-based service applications.The release and processing of real-time vehicle and mobile user trajectory will greatly facilitate people’s daily life and travel.However,if the unprotected vehicles and mobile users’ personal trajectory data are published,it may cause users’ personal sensitive information leakage.The privacy disclosure of trajectory data has become an urgent problem to be solved.With the demand of real-time application,real-time trajectory data release also brings greater security challenges to users’ personal privacy.Nowadays,researchers put forward various solutions based on k-anonymity,suppression method and random perturbation for trajectory privacy-preserving.Based on the research and comparative analysis of existing privacy-preserving mechanism,this paper finally adopts the differential privacy method to achieve real-time trajectory privacy-preserving.Under the previous works’ theoretical basis,aiming at privacy leakage possibly caused by real-time trajectory punishment,this paper extends and improves the w-event differential privacy re-quirement and sample and filter method proposed by predecessors,designs and implements a real-time trajectory differential privacy privacy-preserving mechanism while balances tra-jectory data’ availability and privacy.In view of the existing problems in the real-time trajectory data differential privacy-preserving mechanism,this paper implemented areal-time trajectory data differential privacy-preserving mechanism based on sampling and filtering method and meeting w-event differential privacy requirements.Firstly,in order to ensure the data availability and reduce the query publishing error,this paper implements the filtering module according to the Ensemble Kalman Filter,and on the basis of previous work,the Kalman Filter state equation based on the region transfer matrix is designed to improve the accuracy of the prediction module.At the same time,in order to ensure the computational load and practicability of the mechanism,after comparing the variable length sampling with the fixed length sampling,this paper uses the variable length sampling mechanism to sample the data.In this paper,the sampling interval variation principle based on data change rate is introduced in the sampling module design,and the statistical error method based on PID control is used.Combined with statistical er-ror and data change rate,the sampling interval adjustment calculation method is designed according to the exponential mechanism.In addition,this paper analysises the change of the query sensitivity caused by the temporal correlation of trajectory data,and presents a calcu-lation method based on users’ forward and backward transition probability.Finally,in order to improve the user data privacy in the hotspot region,a regional privacy weight division method based on access frequency and regional neighbor relationship is proposed,and in order to meet the w-event differential privacy requirement,the regional privacy budget allo-cation method based on regional privacy weight and the equal ratio series is adopted in the mechanism.Through analysis and experiment,the real-time trajectory privacy-preserving mechanism realized in this paper guarantees a certain degree of usability while protecting the real-time trajectory data,which has certain practical value.