Research On Differential Privacy Budget Allocation Method For Location Big Data Statistical Publishing

The wide application and iterative update of big data technology have accelerated the pace of human beings entering the era of data and intelligence.Various big data services based on location information provide unprecedented intelligent and personalized services,which bring great convenience and speed to people’s lives.However,the improper use of location big data also causes the leakage of users’ personal privacy data,which poses a huge threat to user information security.The combination of differential privacy model and location big data statistics publishing method can effectively realize the privacy protection of location big data statistics without considering the background knowledge of the attacker.However,the statistical release results of location big data based on differential privacy model introduce irreversible noise,which has an impact on the analysis and application based on accurate data.Therefore,how to further improve the availability of location big data statistics release results under the premise of ensuring user privacy is a major difficulty at present.This thesis mainly studies the differential privacy budget allocation method based on location big data statistics.The reasonable allocation of differential privacy budgets in the statistical release is closely related to the accuracy of the results of the statistical information release of the location.This thesis proposes a method of equal differential privacy budget distribution method released by big data statistical division of big data statistics.First,through the analysis of statistical distribution structures and release errors,it derives the hierarchical distribution formula of equivalent privacy budgets.Then analyze comparison of equivalent budget distribution methods with other privacy budget distribution methods.This method has obvious advantages in the results and overall errors in each layer of privacy budget allocation.The privacy budget increases from the root node to the leaf node according to the method of the equal ratio,which effectively reduces the error of the query area counting value.Through the experiment of the actual location big data set,the equivalence of the differential privacy budget distribution method is significantly better than other privacy budget distribution methods in terms of scope counting accuracy,which helps improve the availability of large data statistics.The traditional privacy budget distribution method does not fully consider the effects of data distribution features on the distribution of privacy budgets,and cannot effectively meet the needs of privacy protection released by data division.Therefore,this thesis proposes a method of dynamic differential privacy budget distribution released by big data statistics.In the tree-type division structure,by considering the number of nodes per layer,the scope counting query error is reduced layer by layer according to the gradient decrease of algorithm ideas.This method first allocates the initial value of privacy budget for the root node,and then distributes the privacy budget according to gradient descent algorithm ideas.If the number of neighboring nodes gradually increases,the current privacy budget remains unchanged,otherwise the privacy budget will be increased.Through the experiment of the actual location big data set,the big data statistical division of the big data statistics of this thesis is superior to other privacy budget allocation methods,which is of great significance to improve the availability of the release of the release of the release data.