差分隐私保护关键技术研究

Differential privacy is a new privacy technology developed in recent years. When the attacker attempts to infer the target to reveal the real information, differential privacy algorithm told the attacker： no matter what queries submitted, regardless of how many times queries submitted, the target is absent or not in this data set, the differences between two queries replied by the algorithm is not too much. The real distribution of data set is hidden by the technology of differential privacy, and is believed to offer strong privacy protection. To my understanding, the issue of the level of differential privacy, as well as how to measure the availability of differential data is still unknown. From this perspective, we first to study the privacy level of differential privacy, and then analyze the measure of differential privacy protection model. Based on this, we can find the best balance point between the utility and the level of privacy protection offered by differential privacy, which provide a reference for the application of differential privacy. In addition, the idea underlying differential privacy will be applied to the privacy of protection in the cloud computing and moving object privacy-aware query.

差分隐私保护是近几年发展起来的一种新的隐私保护技术，当攻击者试图推断出攻击对象的真实信息，差分隐私算法告诉攻击者：无论提交什么样的查询，无论攻击者提交多少次查询，这个攻击对象在或者不在这个数据集，差分隐私算法返回的查询结果差别不大。差分隐私保护技术隐藏了数据集真实的数据分布，提供了很强的隐私保护力度。但是，目前，还不清楚差分隐私到底提供了多强的隐私保护力度以及如何去度量差分隐私的数据可用性。有鉴于此，本项目重点研究差分隐私保护技术的性质及其应用, 拟解决差分隐私保护的力度，同时研究差分隐私保护的可用性度量模型，之后，可以找到差分隐私的数据可用性与隐私保护力度二者之间的最佳平衡点，为差分隐私技术应用提供参考。此外，拟将差分隐私应用到云计算和移动对象的隐私保护查询。

本课题对面向差分隐私保护的关键问题进行深入研究，完成了原定研究计划并取得了以下主要成果：(1)建立了差分隐私保护技术的攻击模型，提出了相应的攻击模型定义，并给出了攻击者成功概率；(2)提出了差分隐私保护技术参数e的选取计算式；(3)研究隐私保护异构匿名化算法；(4)提出了基于置换的数据匿名算法， 5)提出了多力度用户隐私保护技术，5）证明了(e,m)-匿名问题是NP-hard 问题， 6）基于上述工作，本项目共发表论文10篇，其中国际刊物论文2 篇、国际会议论文5篇、国内权威期刊3篇，其中SCI检索2篇，EI检索8篇。培养研究生4名。我们所做的工作能够促进差分隐私保护技术的研究及应用。