Research On Trajectory Differential Privacy Protection Method For LB

With the gradual maturity of positioning technology and the rapid development of mobile communication,Location Based Service(LBS)has been applied widely and deeply,and the collection,analysis and mining of mobile user’s trajectory have become more and more convenient.However,the trajectory contains a lot of sensitive personal information of users such as home address,workplace,consumption capacity,hobbies and so on,and the correlation between any trajectory often reflects a certain social relationship between users.Once the trajectory data is sold illegally or attacked maliciously,it will lead to serious privacy leakage problem of user.Therefore,while enjoying the dividends of LBS,how to protect user’s sensitive personal information and social relationship information associated with trajectory effectively has become a hot research topic in the field of privacy protection.Aiming at the above problem,in-depth research with the help of differential privacy technology has been conducted in this thesis,which is mainly summarized in the following two aspects.(1)The trajectory differential privacy protection method based on noise trajectory similarity tree is proposed to address the problem of personal sensitive information leakage contained in user’s trajectory.Firstly,a similarity measurement between trajectories is proposed.The trajectory is sliced and sequenced simultaneously in both time and space dimensions,and the inter-trajectory similarity is calculated by combining the factor of trajectory shape to improve the accuracy of the similarity metric.Secondly,a noise trajectory similarity tree structure based on differential privacy is designed by combining the above metric results,which improves the resistance to arbitrary background knowledge attack while achieving efficient storage and query of trajectory sequence.In addition,a mutual correlation constraint is applied to the noise trajectory sequence,which achieves their logical and temporal correlation with the noise sequence and the original trajectory sequence to avoid attackers from identifying user’s real trajectory.Finally,experiments show that the method has higher data availability and data privacy,and achieves a better balance between the two.(2)The differential privacy protection method for interrelated trajectory based on Lagrangian optimization is proposed to address the problem of social relationship leakage caused by the correlation between different user’s trajectory.Firstly,an inter-trajectory correlation scoring method is designed.The correlation degree between different trajectoriy is measured by calculating the value of inter-trajectoriy correlation score function,which indirectly quantify the social relationship between users.Secondly,the Laplace distribution scale parameters are optimized by the Lagrange multiplier method,and the normalized set of Laplace distribution scale parameters is obtained.In addition,the optimal Laplace noise set is added to the original trajectory to achieve privacy protection for inter-trajectory correlations.Finally,the analysis of attacker’s prior-posterior knowledge and experiments show that the method protects the inter-trajectory correlations effectively while ensuring high data availability.