一种面向图模型数据的差分隐私保护算法  

A Differential Private Preserving Algorithm for Graph Model Date

作　　者：宋宇宁[1] 丁丽萍[1] 李彦峰[1] 刘雪花[1] SONG Yu-ning;DING Li-ping;LI Yan-feng;LIU Xue-hua(Laboratory of Parallel Software and Computational Science,Institute of Software,Chinese Academy of Sciences,Beijing 100190 China)

机构地区：[1]中国科学院软件研究所并行软件与计算科学实验室,北京100190

出　　处：《自动化技术与应用》2020年第7期86-90,共5页Techniques of Automation and Applications

基　　金：国家重点研发计划基金资助项目(编号2016QY01W0200);国家自然科学基金面上基金资助项目(编号61772507);广州市科技计划项目(编号201802020015);贵阳信息技术研究院(中科院软件所贵阳分部)区块链技术与应用联合实验室开放课题。

摘　　要：在公开的社会网络数据模型研究中,随机图模型是当前流行的针对社会关系数据的统计和研究方法,其数据通常来源于公开的社会网络数据,而差分隐私技术的理论完备性正适合此类信息的收集、分析和发布。本文对差分隐私在图模型数据中的应用做了概述,提出了一种用于随机图模型数据中的差分隐私方法并讨论了其效能、应用场景以及当前的局限性,最后对差分隐私在该领域未来的探索方向做了简要的分析。Among the research models of social network analysis(SNA),random graph model is one of most popular statistical method for social relationship data nowadays.The information of SNA usually comes from public data,while DP with its theoretical completeness of DP is designed to solve privacy problems on conditions such as collection,analysis and publication of this kind of information.This article gives an overview of the application of DP in random graph model data,proposes a DP algorithm for model fitting in exponential random graph model(ERGM)data,and discusses its effectiveness,application scenarios,and also limitations.The future research direction of the field is briefly descripted analyzed.

关 键 词：隐私保护 差分隐私 随机图模型 模型参数估计 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]