Research On Privacy-preserved Methods For Location-based Social Network

The rapid development of location-based social networks,such as Foursquare,has generated a lot of user behavior data related with location every day.Based on these data,LBSNs and other third-party service provider can provide personalized services which have improvedOcxvb user experience.However,these user data,which contains a large amount of privacy information,may endanger users' privacy.Therefore,it is of great importance to study the privacy protection of user behavior data in LBSN.The thesis investigates the LBSN privacy issue when users subscribe third party services from historical data release and online data release.The main contributions are as follows:(1)We propose a two-step choice privacy-preservation method for historical check-in data publishing.The idea of the proposed method is to cluster users based on users' check-in data first Then the method chooses a cluster,which satisfies threshold of privacy and is the nearest to the cluster that the publishing user belong to,and then chooses a user in the chosen cluster whose check-in data satisfies the threshold of privacy preservation with smallest distortion.Finally,the method publishes chosen user's check-in data as the publishing user's.Empirical experiments show that the two-step choices method is not only protecting privacy,it is also having a higher data utility.(2)To resist inference attack more effective,we propose an anatomy and reconstruction privacy-preservation method with information entropy constraint.When a user subscribes a third-party service,the third-party service provider will get access to the users' online query for personalized service.When the LBSN platform shares users' query,it needs to adopt some methods to prevent user from privacy exposed.Based on anatomy and reconstruction method,we further anatomize the query for expanding the request set appropriately,to improve the security of the query set;To solve the problem of excessive communication cost caused by large query set,we propose greed,envelope growth and common ancestors three strategies to get a query set,which fit privacy preservation threshold with less communicate cost.Experiments show the query sets generated by the three strategies we proposed in this thesis can fit the privacy preservation threshold,at same time,the query set is smaller.(3)We propose a location privacy preservation method based on behavioral similarity,to preserve users' location privacy of LBSN users.Our idea is to cluster users based on users'hierarchical data.Upon a user wants to send a query,he will change his location with other users' queries,which sent by users in same cluster.By this way,adversaries are harder to identify users' true location.We also point out that the proposed method may bring about the reduction of service quality and proposes a solution by expanding the size of the user cluster as time goes on.Empirical experiments show that users enjoy a high anonymity and alleviates the risk of that users cannot receive service for a long time.