基于Diffpart的集值型数据动态发布算法

针对集值型数据动态发布中添加噪音量过大、运行时间长、数据可用性低的问题,提出一种基于Diffpart算法的发布算法。该算法首先将数据集按Diffpart算法构造分类树;而后,利用随机抽样法对分类树节点进行抽样,并对抽样节点对应数据统计值应用差分隐私机制添加噪声,非抽样节点对应数据统计值直接发布。最后,随机生成移位数对抽样点进行调整,实现后续数据的动态发布。实验证明动态发布算法在保护性和实用性上均达到了理想效果。

The algorithm of publishing dynamic set-valued data based on Diffpart

An algorithm based on the Diffpart was proposed for publishing the dynamic set-valued data to solve the problem of adding large amount of noise, spending too much time, having low practicality. Firstly, set the data by Diffpart algorithm. Secondly, using fixed sampling algorithm to sample the node, and utilize the differential privacy mechanism to add the noise. Finally, generate the transposing number to adjust the sampling nodes, realizing the dynamic data release. Through the experiment, it is proved that the algorithm achieve a desired results on protective and practicability.