差分隐私保护下的高维数据集发布  

High-dimensional Datasets Publication Under Differential Privacy Protection

作　　者：朱徐亚 ZHU Xuya(School of Computer Science and Engineering,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学计算机科学与工程学院,安徽淮南232001

出　　处：《洛阳理工学院学报（自然科学版）》2022年第2期73-80,共8页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

基　　金：安徽省重大科技专项基金项目(18030901025).

摘　　要：为了解决维度灾难所引起的隐私保护数据发布计算复杂度高、可用性低的问题,提出基于差分隐私采样机制和贝叶斯网络的DPSM-Bayes算法。利用贝叶斯网络模型,将高维联合概率分布转化为多个低维边缘概率分布,结合差分隐私采样机制和更适合高维概率分布加噪的IMLaplace机制,生成可用性更高的高维合成数据集。实验结果证明,在提供相同差分隐私保护的前提下,DPSM-Bayes算法能够有效地处理高维数据集的发布问题,与现有的方法相比发布的数据集具有更高的质量和可用性。In order to solve the problems of high computational complexity and low availability of privacy protection data release caused by dimensional disasters,a DPSM-Bayes algorithm is proposes based on differential privacy sampling mechanism and Bayesian network.The algorithm uses Bayesian network model to convert high-dimensional joint probability distribution into multiple low-dimensional edge probability distributions,combined with differential privacy sampling mechanism and the IMLaplace mechanism proposed in this paper.The results have proved that under the premise of providing the same differential privacy protection,the DPSM-Bayes algorithm can effectively deal with the release of high-dimensional datasets,and compared with existing methods,the released datasets have higher quality and availability.

关 键 词：差分隐私 数据发布 高维数据 贝叶斯网络 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]