Differential Privacy-based Data Publishing Method

With the widespread use of mobile positioning devices such as mobile phones,the trajectory data of moving objects is more and more likely to be generated and collected.Analyzing and mining these trajectory data can get a lot of useful information.However,how to protect the sensitive information of the moving object contained in the trajectory data from being leaked is a very important practical problem,and it is also a research hotspot in the database field.The traditional trajectory data privacy protection method represented by k-anonymity is to uniformly divide trajectory data that satisfies disjoint privacy constraints into the same group,and then add statistics on the records in the same group to perform statistical release.Traditional trajectory data privacy protection method assumes that the attacker does not grasp the background knowledge related to the trajectory data.In the Internet age,the attacker can obtain enough background knowledge through many channels.Therefore,it is difficult to provide a strong privacy protection for traditional track data privacy protection methods.Differential privacy is a major method of privacy protection currently for trajectory data.Differential privacy has an information-theoretic security.It adds noise to the data through perturbation,making the information entropy of the entire data almost unchanged.Even if the attacker knows part of the data,he cannot guess the specific individual information in the data.This paper focuses on the research of real-time publishing of trajectory data based on differential privacy.The main research contents and research results are summarized as follows:For the trajectory data of a certain time,a histogram distribution algorithm APG based on differential privacy is proposed.APG first uses the ranking algorithm based on the exponential mechanism,and then combines the AP clustering algorithm to adaptively cluster the grouped results,and finally adds noise to the grouped results.For the trajectory data whose length of the moving object is L,a dynamic privacy budget allocation algorithm RTPM is proposed.The RTPM uses an exponential decay mechanism to rationally allocate the current privacy budget,and reserves a corresponding size of privacy budget for future trajectory data.The innovation of this article has the following points:A sorting algorithm based on exponential mechanism is proposed.In combination with AP clustering algorithm,adaptive grouping is achieved through MSE index and clustering factor ? to obtain the optimal grouping strategy.On the basis of satisfying the ?-differential privacy requirement,the approximation error AE produced by the group mean value and the Laplacian error LE caused by adding Laplacian noise are balanced,which effectively reduces the error MSE of the published data and improves the data.Availability.A dynamic privacy budget allocation algorithm was proposed to allocate the privacy budget for each moment of track data autonomously,it improves the utilization of the privacy budget,and better protects the privacy of the trajectory data whose length of the mobile object is L.The dynamic privacy budget allocation algorithm is dynamically allocated for each moving object,a trajectory of length L,and the sum of the L-trajectory privacy budgets of each moving object needs to be less than or equal to the total privacy budget.The algorithm allocates half of the unused privacy budget for each moving object at the current moment,and reserves the other half of the privacy budget for the future trajectory data of the moving object.