本地化差分隐私研究综述

大数据时代信息技术不断发展，个人信息的隐私问题越来越受到关注，如何在数据发布和分析的同时保证其中的个人敏感信息不被泄露是当前面临的重大挑战.中心化差分隐私保护技术建立在可信第三方数据收集者的假设基础上，然而该假设在现实中不一定成立.基于此提出的本地化差分隐私作为一种新的隐私保护模型，具有强隐私保护性，不仅可以抵御具有任意背景知识的攻击者，而且能够防止来自不可信第三方的隐私攻击，对敏感信息提供了更全面的保护.介绍了本地化差分隐私的原理与特性，总结和归纳了该技术的当前研究工作，重点阐述了该技术的研究热点：本地化差分隐私下的频数统计、均值统计以及满足本地化差分隐私的扰动机制设计.在对已有技术深入对比分析的基础上，指出了本地化差分隐私保护技术的未来研究挑战.

Survey on Local Differential Privacy

With the development of information technology in the big data era, there has been a a growing concern for privacy of personal information. Privacy preserving is a key challenge when releasing and analyzing data. Centralized differential privacy is based on the assumption of a trustworthy data collector; however, it is actually a bit difficult to realize in practice. To this end, local differential privacy has emerged as a new model for privacy preserving with strong privacy guarantees. By resisting adversaries with any background knowledge and preventing attacks from untrustworthy data collector, local differential privacy can protect private information thoroughly. Starting with an introduction to the mechanisms and properties, this paper surveys the state of the art of local differential privacy, focusing on the frequency estimation, mean value estimation and the design of perturbation model. Following a comprehensive comparision and analysis of existing techniques, further research challenges are put forward.