摘要
针对满足差分隐私的高维数据发布问题,我们提出一种基于隐树模型的满足差分隐私的高维数据发布算法.该算法由隐变量生成、隐树结构学习、隐树参数学习和数据生成四个阶段组成.特别地,在该算法中,为了在对显变量进行分组并生成隐变量的过程中保护隐私,我们提出一种满足差分隐私的隐变量生成方法.此外,为了在构建隐树的过程中保护隐私,我们提出了一种满足差分隐私的隐树模型结构学习方法.分析结果表明本文提出的算法满足ε-差分隐私.实验结果表明,与现有算法相比,所提出的算法可以获得更好的数据效用.
To solve the differentially private high-dimensional data publishing problem,we present a latent tree model based differentially private data publishing algorithm.This algorithm consists of four phases:latent variable generation,latent tree structure learning,latent tree parameter learning and data generation.In particular,in this algorithm,to privately group the observable variables and generate the latent variables,we propose a differentially private latent variable generation method.Moreover,to privately construct the structure of the latent tree,we propose a differentially private structure learning method for the latent tree construction.Formal analysis shows that the proposed algorithm satisfies ε-differential privacy.Experimental results demonstrate that the proposed algorithm can obtain better data utility than existing algorithms.
作者
苏炜航
程祥
SU Wei-hang1, CHENG Xiang2(1 The High School Affiliated Renmin University of China,Beijing 100080 ,China;2 State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, Chin)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第4期681-685,共5页
Journal of Chinese Computer Systems
关键词
隐树模型
高维数据发布
差分隐私
隐私保护
latent tree model
high-dimensional data publishing
differential privacy
privacy protection