自适应差分隐私预算分配策略的直方图发布算法

差分隐私直方图发布中，隐私预算涉及到噪声添加的强度，直接影响到直方图发布的数据可用性，如何合理地进行隐私预算的分配是直方图发布算法面临的一大挑战。提出了一种自适应的隐私预算分配策略的直方图发布算法APB（adaptive privacy budget allocation），首先通过分析分组前后引入的噪声误差和重构误差，建立了隐私预算分配权重的优化模型，得到最优分配权重和分组大小以及分组个数之间关系；然后基于优化模型和贪心分组的思想，提出了自适应的隐私预算分配策略，可以更好地均衡噪声误差和重构误差，提高发布数据的可用性。实验结果表明，基于自适应的隐私预算分配策略的直方图发布算法可用性高于同类算法。

Histogram publishing algorithm based on adaptive privacy budget allocation strategy under differential privacy

In differential privacy histogram publication the privacy budget involved the intensity of noise adding, which had a direct impact on the availability of data publication. Therefore, how to allocate and optimize the privacy budget was a challenge to the histogram publishing algorithm. This paper proposed a histogram publishing algorithm based on adaptive privacy budget allocation strategy, called APB. In order to balance the noise error and reconstruction error efficiently and get the relationship between the privacy budget allocation weight and the group size and number of group, APB established an optimization model of privacy budget allocation weight by analyzing the noise introduced before and after the grouping step. In order to boost the accuracy of the released histogram, this paper employed optimization model and greedy grouping to propose an adaptive privacy budget allocation strategy. Experimental results show that the proposed method of histogram publication based on adaptive privacy budget allocation strategy is more available than similar algorithms.