Research On Spatial Partition Data Publishing Method Based On Differential Privacy Protection

With the rapid development of large data,the demand for the collection of spatial data is becoming increasing.The distribution of data is an important factor in the spatial partitions,to meet the requirements of differential privacy.When the interval is too coars,the uneven distribution of data will cause the classification error increases,when the interval size selection carefully,similar data distribution will cause noise error accumulation.In order to solve the adaptability of interval division and data distribution,this paper proposes a hierarchical optimization method for logical grid based on differential privacy.The method uses bottom-up ideas,first of all the data sets are fine-grained grids,according to data distribution similarity is merged to form relatively coarse-grained grid,the new area to form the two division according to the calculation of the contribution rate of the final query,constitute the three hierarchy model.The concept of logical grid unit,solve the grid unit is too small,the noise grid large number of elements due to accumulation of problems.The concept of logical grid domain is proposed,forming a set of logical grid units,constructing hierarchical models of logical grid domains,and using consistency constraints to process internal nodes,which improves the utility of queries.In order to reduce the time complexity of the HOLG-DP classification method,this paper proposes the use of Huffman tree to optimize the processing hierarchy tree generated by HOLG-DP,through the construction of Huffman tree hierarchy tree of the same type,the weighted path length distribution hierarchy tree large area in the same type in short,in order to realize the data domain to be divided on the issue to guarantee the query effectiveness at the same time,effectively reduce the query response time division algorithm.To verify the feasibility of using HOLG-DP method to optimize the Huffman tree,through changes in the actual data sets were observed under different thresholds of relative error;the accuracy comparison between HOLG-DP and different classification methods in the privacy protection budget under the same query results,and through the comparison of the different parameters under different privacy protection division method operation response time.Two groups of experiments can see that the relative error has been significantly reduced,the effectiveness of published query results in a certain scale of query area has been significantly improved,and the latter can effectively reduce the running response time.