面向社交网络分析的差分隐私保护研究综述  

A survey on differential privacy research for social network analysis

作　　者：王俊丽[1] 管敏[1] 魏绍臣 

机构地区：[1]同济大学CAD研究中心,上海201804

出　　处：《高技术通讯》2015年第3期239-248,共10页Chinese High Technology Letters

基　　金：国家自然科学基金(61105047);港澳台科技合作项目(2013DFM10100);上海市科委项目(14JC1405800);国家科技支撑计划(2012BAF12B11)资助项目

摘　　要：阐述了数据的差分隐私保护概念,给出了差分隐私保护模型,从理论上描述了其噪声机制和组合性质,着重进述了差分隐私保护模型在社交网络发布数据隐私保护上的应用及发展,给出了差分隐私保护应用于度分布查询、子图计数、聚类系数计算、边权重计算等社交网络分析技术的实验结果。分析发现,研究差分隐私保护应重点考虑隐私预算和噪声机制,隐私预算决定了隐私保护强度,噪声机制决定了查询准确性;探讨差分隐私保护在社交网络领域的应用,是一个重要的研究方向。The concept of differential privacy protection of data is interpreted. The differential privacy model, and its noi- sing mechanism and combination properties, are theoretically described and anlayzed. The application of the differ- ential privacy model to social network data＇ s privacy protection and its development are emphatically reviewed with a rigorous, quantitative representation, and the experimental results of differential privacy applications to the social network analysis techniques of degree distribution inquiry, Subgraph counting, clustering coefficient computation and edge weight computing are given. It is concluded from analysis that the privacy budget and the noising mecha- nism are the main factors to differential privacy （the former determines the privacy protection intensity, while the latter determines the inquiring accuracy）, and exploring the application of the differential privacy protection to the social network field is the main future research direction.

关 键 词：差分隐私保护 社交网络分析 图挖掘 统计方法 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]