局部差分隐私的新型实现方法

New Implementation Method of Local Differential Privacy

有效的隐私保护数据发布解决方案之一是局部差分隐私,随机响应是实现这种隐私保护模型的有效方式。对基于二次扰动的局部差分隐私实现方法进行了研究。为衡量D和D'的离散程度,在计算原始数据集和扰动数据集的分布均值和方差的基础上实验验证了D和D'间的KL-散度。实验结果表明本文所采用的二次扰动方法可以带来较小的效用损失。

One of the effective privacy protection data publishing solutions is local differential privacy, which is an effective way to implement this privacy protection model. This paper proposes a local differential privacy implementation method based on secondary perturbation. In order to measure the degree of dispersion of D and D', the KL-divergence between D and D' is experimentally verified on the basis of calculating the mean and variance of the distribution of the original dataset and the perturbed dataset. The experimental results show that the secondary perturbation method used in this paper can bring less utility loss.