• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 计算机网络与信息安全 • 上一篇    下一篇

本地差分隐私保护及其应用

高志强,崔翛龙,周沙,袁琛   

  1. (武警工程大学乌鲁木齐校区,新疆 乌鲁木齐 830049)
  • 收稿日期:2017-11-02 修回日期:2018-02-15 出版日期:2018-06-25 发布日期:2018-06-25
  • 基金资助:

    国家自然科学基金(U1603261);新疆维吾尔自治区自然科学基金(2016D01A080)

Local differential privacy protection and its applications

GAO Zhiqiang,CUI Xiaolong,ZHOU Sha,YUAN Chen   

  1. (Urumqi Campus,Engineering University of PAP,Urumqi 830049,China)
  • Received:2017-11-02 Revised:2018-02-15 Online:2018-06-25 Published:2018-06-25

摘要:

隐私保护问题已成为信息安全领域研究的重点方向。差分隐私从2006年提出至今一直受到理论界的推崇,而近年来在产业界众包模式下的本地差分隐私受到了极大关注。分析了本地差分隐私模型相对于经典差分隐私模型的演进与应用场景,从理论研究和工程实践角度,对本地差分隐私基础理论及其在数据收集与数据分析中的应用研究进行综述。在数据收集方面,介绍了本地差分隐私的主要研究和应用成果,并着重从差分隐私的角度对这些方法进行了分析比较。在数据分析方面,阐述了本地差分隐私在编码、解码以及在统计学角度的实现和分析方式,并从理论上对这些算法进行推导分析。最后,在对已有技术深入对比分析的基础上,总结出了本地差分隐私技术面临的挑战和研究方向。
 
 

关键词: 差分隐私, 数据发布, 数据挖掘, 机器学习, 众包, 隐私保护

Abstract:

Privacy protection has become the focus of information security research. Differential privacy has been highly recommended by the theoretical community since 2006. In recent years, the local differential privacy in the industry crowdsourcing model has attracted much attention. We analyze the local differential privacy model from the perspective of theoretical research and engineering practice, and summarize the applications of local differential privacy theory in data collection and data analysis. In the aspect of data collection, the main research and application results of local differential privacy are introduced, and the methods are analyzed and compared from the perspective of differential privacy. In the aspect of data analysis, the implementation and analysis of local differential privacy in encoding, decoding and statistics are discussed, and these algorithms are analyzed theoretically. Finally, on the basis of indepth comparison and analysis of existing technologies, the challenges and research directions of local differential privacy techniques are summarized.
 

Key words: differential privacy, data publishing, data mining, machine learning, crowdsourcing, privacy protection