A Trajectory Privacy Protection Method Based On Differential Privacy Mechanism

In recent years,the protection of mobile users' growing track data before data mining has become the mainstream research direction in the field of track privacy protection.In this way,a differential privacy track publishing method based on the idea of track aggregation has been developed.Although this algorithm model provides a novel algorithm idea for differential privacy trajectory protection,there are still many shortcomings.For example,K-means clustering is selected in the algorithm,so the number of clustering depends on artificial selection.In the data generalization stage,the temporal dimension is ignored and only the position point is generalized.When data noise is added to the data set,the usability of released data is poor.To solve the above problems,this paper proposes a spatiotemporal multi-dimensional density feature generalization trajectory publishing method MGR based on differential privacy.The main research contents are as follows:Firstly,in view of the k-means clustering is not applicable to extract the trajectory characteristics of multiple attribute data,this paper proposes a Adaptive clustering algorithm based on Density characteristic value(the Adaptive X-means-based on Density Eigenvalue,AXDE)and puts forward the concept of Density of eigenvalues,AXDE by calculating data sets the Density of sample points characteristic value can adaptively choose the number and position of the initial clustering center,make the clustering results more conform to the requirements of the clustering of trajectory data set.Secondly,in view of the present model of algorithm based on trajectory clustering space for only track data generalization,and ignores the problem of time generalization,this paper puts forward the MGR algorithm using AXDE respectively from the two dimensions of temporal and spatial trajectory data clustering generalization,always keep the same set of trajectory have enough space,trajectory set the location of the generalization result more primitive characteristics,in order to promote the availability of published data sets to provide reliable premise.Thirdly,for trajectory data set in add uniform noise and phase noise problem,put forward a numerical weight meter based on the epsilon?w-differential privacy model,the model based on the size of the data count weight computing privacy,using Laplace mechanism under the established differential privacy budget flexibility to add noise,make the data set of query results relative error is under control,improve availability algorithm.Finally,the differential privacy trajectory privacy protection method proposed in this paper is verified through two groups of experiments.First of all,AXDE is compared with the existing clustering algorithm to verify the availability of the AXDE clustering algorithm in MGR.And then,the MGR and DPG methods are compared in the real trajectory data set.By comparing the error analysis of counting query results,it is concluded that the differential privacy trajectory publishing method proposed in this paper has more data availability.