Latent Tree Model Based Differentially Private Highdimension Data Publishing Algorithm
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,China）
Abstract：To solve the differentially private highdimensional 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 εdifferential privacy.Experimental results demonstrate that the proposed algorithm can obtain better data utility than existing algorithms.
苏炜航,程祥. 一种基于隐树模型的满足差分隐私的高维数据发布算法[J]. 小型微型计算机系统, 2018, 39(4): 681-685.
SU Wei-hang,CHENG Xiang. Latent Tree Model Based Differentially Private Highdimension Data Publishing Algorithm. Journal of Chinese Computer Systems, 2018, 39(4): 681-685.
2018, Vol. 39 
