Optimization And Application Of Privacy Budget In Differential Privacy Protection

In data privacy preservation,we demand not only to maintain data privacy but also to improve the utility of data.The allocation of privacy budget in differential privacy determines how much noise is added to the true results,that influences the security and utility of data.As the total number of queries gains or is even unlimited in some application scenarios,the amount of added noise which reduces the data utility would increase rapidly.To solve the allocation and optimization problems of privacy budget with limited queries in practical,we propose several novel schemes for differential privacy by using series to allocate privacy budget.The privacy budget?is generalized by expressing in the form of series,and the i-th budget?_i is the i-th component of the series expression,such that the sum of privacy budgets does not exceed?.It can reduce the incremental speed of noise added.Three types of series,Taylor,p series and special series,are evaluated for allocating privacy budget.They can be used to form some privacy preserving methods with unlimited processing.Moreover,for some applications,data processing is limited.Some optimized allocations of privacy budget are proposed and analyzed for these cases.The theoretical analysis and experimental results show that the series-based methods satisfy the?-differential privacy,and that the p-series and special series methods add less noise than the bisection method.Therefore,the utility of data is enhanced.