Abstract
In this paper, we develop a novel strategy for the privacy budget allocation on answering a batch of queries for statistical databases under differential privacy framework. Under such a strategy, the noisy results are more meaningful and achieve better utility of the dataset. In particular, we first formulate the privacy allocation as an optimization problem. Then derive explicit approximation of the relationships among privacy budget, dataset size and confidence interval. Based on the derived formulas, one can automatically determine optimal privacy budget allocation for batch queries with the given accuracy requirements. Extensive experiments across a synthetic dataset and a real dataset are conducted to demonstrate the effectiveness of the proposed approach.