Research On Trajectory Semantic Privacy Protection Methods Based On Trajectory Segment Clustering In Road Networks

In recent years,the popularity of mobile smart devices has brought great convenience to people's lives.And service providers have accumulated a large number of users' trajectory data.Because of the rich space-time information,the publication of trajectory data set is of great value to intelligent transportation,scientific research,business analysis and government management.However,if the service providers directly publish the original data,the users' trajectory privacy will be leaked.And the attacker will infer the sensitive privacy such as interests,behavior patterns and home address.In order to publish data safely,scholars have proposed many trajectory privacy preserving methods.While many methods ignore the semantic attributes.Even if the anonymous trajectory data is published,the semantic privacy of the trajectory will still be disclosed.In this paper,we introduce the(k,l,t)-anonymity model based on the generalization method.In order to solve the problems of variable trajectory movement state,coarse semantic security granularity and topological constraints of road networks,we design a trajectory semantic privacy protection method based on trajectory segment clustering in road networks.The method is divided into six stages.(1)The original trajectory data is preprocessed.The trajectory points are converted to the road network vertices to reduce the possibility of trajectory restoration.(2)Based on the MDL principle,the trajectory partition(TP)algorithm is proposed.And the anonymous unit is refined from the whole trajectory to the local segment.(3)The clustering(CBRND)algorithm based on segment space-time distance is proposed to cluster trajectory segments to form multiple segment classes.(4)The local region partition(LRP)algorithm is proposed.And we calculate the semantic location distribution of the local region of the segment class.(5)Two anonymous algorithms are proposed to construct semantically secure anonymous sets.The trajectory segment merging(TSM)algorithm takes the semantic location distribution of the city as the semantic reference and merges the trajectory segments to meet the semantic requirement first.Then it satisfies the k-anonymity to generate the anonymous set.While the trajectory segment generalization(TSG)algorithm first selects k different user's trajectory segments to satisfy k-anonymity.And then it takes the semantic location distribution of the local region of the segment class as the semantic reference to generate a refined semantic secure anonymous set.(6)The anonymous sets are published safely.We use a real map to conduct experiments in two stages.The first stage is the result analysis of clustering algorithms.The results show that the CBRND algorithm has high contour coefficient and good clustering effect.The difference of semantic location distribution between local region of segment class and city shows that the semantic refinement is reasonable.In the second stage,we compare the anonymity algorithms from privacy protection and data quality.The results show that the TSG algorithm has higher success rate of anonymity,weaker semantic exposure,smaller average time gap and smaller average space gap.