Research On The Improvement Of Geospatial Data Publishing Method Of Mobile Social Network Based On Differential Privacy

With the development of science and technology,mobile devices and locationaware technology was combined.It spawned many location-based applications and services,these applications generate a large amount of location-based data in an increasingly information accumulation.Location-based data can reflect the behavior of individuals.This may lead to the disclosure of user's privacy during data publishing.So we need a mechanism for the protection of privacy in data publication.Differential privacy,as a new type of privacy protection framework in the aspect of data obtained the rapid development and widely used,has achieved many research results.However,Two-dimensional geospatial data processing still exist the following problems to be solved.The current data publication method based on grid is simple and effective.But in the equalization error and noise error even assume there is still large room for improvement.Location-based data of mobile social networks often have uneven distribution characteristics.The current differential privacy.published algorithm can not make good use of the data distribution.Discover a huge error in the uneven distribution of data,affecting the availability of published data.For the above problems,it studies the following three aspects in this thesis:(1)This thesis analyzes the data distribution method based on adaptive grid partitioning,which generates a large number of cells at the time of sparse data uneven distribution.For the problem,This thesis proposes a method of adaptive grid partition based on post-process.The core idea of algorithm is,in order to reduce the number of cells and the generation of noise error,to choose a sparse k adjacent rows of cells were combined,then add nosie and get published data sets.Finally,the case of a large number of experiments to verify the post-processing method in an uneven distribution of data can greatly reduced in the sparse number of cells,thereby reduced the noise error and improved the availability of published data.(2)On the basis of the last work,in order to further balance the relationship between noise error and hypothesis error,in this thesis,a data distribution method has been proposed based on the bottom-up cell merging.Firstly,we conduct a fine uniform meshing,and then merge bottom-up based on adjacent cell density,so release data set was been obtained by adding noise in the consolidated block.Finally,the proposed algorithm,adaptive mesh and its improved algorithms are compared,and superiority is verified of the bottom-up merging method in reducing the error of the query,especially in the case of sparse data.However,the data distribution method needs to be improved in the query efficiency.(3)In order to improve query performance,this paper published data privacy differential method based on mixed tree structure.The method used in the overall division is kd-tree index structure,and in the leaf node is divided using a grid structure.So the difference tree structure based on mixed data privacy publish has the characteristics of both a high kd-tree index structure query efficiency and the use of a grid structure equalization noise error and error uniform assumptions,improve the availability of the published data.Finally,the experiment verifies that differential privacy data distribution method based on mixed tree structure do better in query efficiency comparing with the method of data grid structure.