Research On Trajectory Privacy Protection Technology Based On Hidden Markov Model And Differential Privac

In recent years,with the quick development of smartphones and global positioning system technology,location-based services(LBS)have found wider applications.However,since mobile users have to share the real-time location or location sequences with LBS service providers,query requests can be used to infer sensitive information,such as home/workplace,physical health information,and interpersonal relationships.Therefore,serious privacypreserving risks exist in the LBS-based applications.Considering the preferences of users,this paper considers two LBS application scenarios: i).the high service quality is required and users will send their real location,such as driving navigation;ii).users want to ensure their location privacy with a certain degree of deterioration in the quality of service(Qo S),such as checking the local weather.This paper focuses on these two types of application scenarios of LBS privacy-preserving and obtains the following research results:(1)Aiming at the potential security risks in continuous queries in which users send real locations and adopt k-anonymity privacy protection mechanism,a hybrid spatiotemporal attack algorithm based on the hidden Markov model(HMM)is proposed.The algorithm can not only mine the real location of the user but also find the user’s movement trajectory.Additionally,in order to measure the leakage risk of the user’s movement trajectory,the definition of trajectory disclosure probability is proposed.Simulation results show that the proposed algorithm can achieve a higher trajectory disclosure probability than other algorithms.(2)Aiming at the potential security risks of using the k-anonymity mechanism to send real location for trajectory privacy protection,an advanced k-anonymity algorithm based on HMM is proposed.First,the algorithm achieves indistinguishability of all locations in the same anonymity set.Then,dummy trajectories are constructed that cannot be distinguished from real trajectories by the adversary.In addition,a greedy algorithm is proposed to select the desired dummy locations with a low complexity.Simulation results show that the proposed greedy algorithm can effectively reduce the computational complexity without significantly reducing the accuracy of the results,and the proposed defense algorithm can effectively resist the new attack method of using HMM to infer the real trajectory of the user.(3)When the same user does not send the real location and frequently passes the same path,aiming at the problem that the distribution of dummy locations will expose the real location,a trajectory privacy protection mechanism based on differential privacy is designed.This mechanism uses R-tree to store generated dummy locations that meet differential privacy requirements.When initiating a query,the number of dummy locations within a certain range is limited by first searching the R-tree for the existence of eligible dummy locations instead of directly generating the new location.The simulation results show that the proposed privacypreserving mechanism can reduce the overhead of privacy protection by sacrificing a certain Qo S,i.e.,it can reduce the number of dummy locations by using historical dummy locations rather than newly generated onesFinally,this paper summarizes the research content and prospects the future work.