Research On Privacy Protection Methods Of Spatial Data

With the rapid development of information technology and the rapid spread of global positioning systems,the query and mining technologies for spatial data have become more and more mature.As one of the most common data types,spatial data is widely used in geographic information systems,multimedia systems,aerospace systems,etc.due to its large amount of data,continuous data integration and constant changes over time.However,for the query mining of spatial data,while obtaining favorable information,the accompanying privacy security issues cannot be ignored.Major events such as the recent Google search for stealing user location privacy data have become an important issue for people around the world to discuss and continue to pay attention to.As a new type of privacy protection model,Differential Privacy is widely used in data query and rigorously defines the background knowledge of attackers,has a rigorous statistical model,and provides quantitative mathematical analysis and proof.In data mining,there are relatively few academic studies on the combination of spatial data and differential privacy at home and abroad.Therefore,this paper aims to solve the privacy security problem of spatial data query and mining by using differential privacy technology,and to improve the data availability while protecting the query and mining algorithms.In this paper,the index of spatial data is used to represent the structural quadtree,and the privacy problem existing in the query and mining algorithm of quadtree is studied.The concept of range counting query and quadtree index distance algorithm of quadtree is combined with different concepts.The differential privacy budget allocation method is designed.The query-oriented quadtree differential privacy hybrid decomposition algorithm and the mining-oriented differential privacy quadtree density clustering algorithm are designed.The main work of this paper is as follows(1)The problem of privacy leakage risks that may exist during spatial data query and mining This paper analyzes the advantages and disadvantages of typical privacy protection methods,and studies the combination of differential privacy and spatial data algorithms based on quadtree structure to achieve a balance between privacy protection and data utility(2)Aiming at the privacy security problem brought by spatial data query,the hierarchical region is firstly divided into sparse and dense regions.The quadtree-based optimization algorithm is proposed for dense regions and based on threshold regions at different levels.A differential privacy budget allocation corresponding to it is adopted.Since the quadtree optimization algorithm is affected by the recursion of input parameters,the form of deviation counting is proposed,which reduces the dependence of input parameters and improves the accuracy of the algorithm.The deeper the recursion depth of the quadtree,the more sensitive the data is.Therefore,the protection of sensitive data is further enhanced by using a geometrically allocated privacy budget for quadtrees of different depths in dense areas.According to the real datasets of the two social networks,the relative error analysis shows that the proposed algorithm is more accurate than the AG algorithm,the UG algorithm and the quadtree algorithm with recursive depth under unified budget allocation.Algorithm also plays a good role in protecting privacy.(3)Aiming at the problem of the error caused by the inaccurate measurement distance and the privacy and security caused by the leakage of distance information in the process of spatial location data mining,a differential privacy based quadtree index distance algorithm based on dimension partitioning is proposed.Due to the large influence of input parameters,an improved DBScan density clustering method is proposed on the quadtree index distance algorithm.The deeper the recursion depth of the quadtree means the more sensitive the data,so the Fibonacci budget allocation method for the deeper data nodes further strengthens the protection of sensitive data.Using three sets of real data sets and a set of synthetic data sets,the F-measure Index and Calinski-Harabasz index analysis show that the proposed algorithm is more improved than the differential privacy-based DBScan density clustering algorithm while satisfying the data privacy security.