Study On Key Technologies Of Location-Privacy Protection In Mobile Application

Technologies such as Mobile Communications,Cloud Computing,and the Internet of Things enable the interconnection and information sharing of everything.Mobile applications are widely deployed in smart cities,military defense,government affairs,commerce,and other fields.Mobile terminals' locations and trajectories are the main factors to mobile applications for providing more accurate and efficient services.The logical inference,based on attribute associations such as location semantics,spatiotemporal correlation,and crowdsourcing elements,brings the leakage risk to the users' sensitive information such as their mobile trajectories and identities.Location privacy protection is facing severe challenges in mobile applications.User's mobile behavioral patterns are time-varying.The lack of time-varying mobility modeling makes the attack-simulating privacy-preserving mechanism unable to resist the logical inferential attacks which are based on the spatiotemporal correlation hidden in users' mobilities.The solutions based on the noised real trajectory have the risk of exposing users' actual mobile trajectory,due to the data leakage incidents and the diversified attribute-association logical inferences.Location privacy protection is closely related to mobile semantics.Existing personalized privacy-preserving techniques,based on customized parameters or single semantic elements,lack sufficient mobile-semantic description between the user and the location.It results in the inability to achieve fine-grained differential privacy protection and the phenomenon of over-protection or under-protection occurs.In the continual mobile scenario,the mobile application service platform and users continuously interact with location information.Location-related service queries and service responses directly threaten the users' location privacy security,and there is no research involving eliminating the privacy risks caused by the service responses.For the characteristics and requirements of time-varying mobility modeling,actual trajectory concealment,differential privacy preservation,and continual location-information interaction,this dissertation conducts researches on the key technologies of location privacy protection in mobile applications and has achieved the following main contributions.(1).Aiming at the problems of time-varying mobility modeling and the actual trajectory concealment,we propose a mobility-aware trajectory prediction solution.Based on the time-partitioning concept,we offer a time-related Markov spatial transfer model,STMarkov.It overcomes the limitation that traditional Markov cannot model the user's time-varying mobile mode.Based on the STMarkov mobile model,we design the user's traveling trajectory prediction framework by making use of the user's time-related low-frequency-visited locations,for replacing the trajectory synthesis method based on the noised actual trajectory and concealing the user's actual trajectories.According to the trajectory prediction framework,we design the mobility-aware prediction solution,resisting the inferential attacks with the user's high-frequency visited locations.It makes up for the lack of time-varying mobility modeling in the privacy-preserving techniques and solves the contradiction between the actual trajectory concealment and participation in the application services.Extensive experiments show that the average accuracy of STMarkov's steady-state distribution reaches84.7%.In the predicted trajectory,the anonymity sets in corresponding time partitions up to 96% can nearly fully cover the user's locations most likely to visit.The high prediction accuracy not only hides the actual trajectory,but also guarantees the quality of service of the application.(2).Aiming at differential privacy-preserving requirements,we propose a Mobile Semantic-aware personalized solution,MSP.According to locations' semantic attributes and the user's staying durations,we represent the user's access roles and perform the user's mobility-related clustering on the location set,forming a hierarchical semantic structure with different granularities.Based on personal access to the location,a dedicated approach is proposed to evaluate the location's privacy sensitivity.Based on the user's access role at the location and the location's privacy sensitivity,we formalized the mobile semantic concept between the user and the location,characterizing the differential privacy-preserving requirements from the user and the location respectively.Based on the feature fusion of mobile semantic elements,we propose an adaptive differential semantic generalization mechanism for the user to participate in the application with different geographic locations which have the same or similar mobile semantics.It realizes privacy preservation from the two dimensions of geography and semantics.Extensive experiments show that MSP reconstructs the trajectory within 500 meters and a direction offset of 21.5 degrees on average for each visited position and preserves about 70% of the mobile semantics in the reconstructed trajectory,balancing the tradeoff between location privacy protection and data availability effectively.(3).Aiming at the privacy risks caused by continual location-information interaction in mobile scenarios,we propose a mobility-aware secure differentially private solution,Con Crowd-DP,achieving the secure crowdsourcing application with bidirectional privacy preservation for task request and acceptance.During the service query stage,we generate the user's time-related probability distribution on location-set and design the differentially private method to synthesize perturbed positions for the user to participate in the application,achieving the privacy-preserving service query.In the task acceptance stage,based on the K-norm DP,Bayesian posterior theorem,and STMarkov,we propose an update method of the user's time-related probability distribution,eliminating the influence of accepted crowdsourced tasks and cross-period migration.It solves the secure problem of differential privacy protection for the subsequent task request due to the continuous association of accepted tasks.Extensive experiments show that Con Crowd-DP generates indistinguishable perturbed locations that satisfy the Differential Privacy(DP)principle from the anonymity sets with a degree greater than 10,within 3 km on average,enabling the user to participate in crowdsourcing securely and continually.The related contributions of this dissertation support mobile application scenarios,such as mobility-perception-based attack simulation,differential privacy protection,secure and continual location-interaction,and actual trajectory concealment.It enhances the universality of mobile applications based on the privacy-preserving profiles,which have significant theoretical and practical meanings for their expansion and deployment in different fields.