差分隐私保护研究综述

Survey of research on differential privacy

差分隐私保护通过添加噪声使数据失真,从而起到保护隐私的目的,对于一个严格定义下的攻击模型,其具有添加噪声少、隐私泄露风险低的优点。介绍了差分隐私保护的理论基础和最新研究进展,详细阐述了分类、聚类等差分隐私学习方法的最新研究情况,介绍了一个差分隐私保护的应用框架PINQ(privacy integratedqueries),并对未来的研究发展方向进行了展望。

Differential privacy approach makes data distortion to preserve privacy by means of adding noise.To a rigorous defined attacking model,differential privacy ensures that adding little amount of noise have a low risk of privacy disclosure.This paper surveyed the definition of differential privacy,showed the newest results in research,introduced algorithms of classify,clustering on differentially private learning,and presented a differential privacy application framework PINQ（privacy integrated queries）.Finally,this paper discussed directions for future research