基于差分隐私保护模型的交互式社交网络分析技术研究

Over the past few years, the proliferation of online social networking services has created numerous amounts of social network data. While these data are valuable to sociologists, economists, data-mining researchers and many others, their containing sensitive information of individuals have aroused serious privacy concerns, which have then quickly led into a growing body of efforts in the literature known as privacy-preserving social network analysis (PPSNA). Despite of the notable achievements that have been made, research in this area is still in its infancy. In this project, we tried to impulse the area by making the following contributions. First, we will propose the Seamless Privacy (SP) framework for privacy-preserving interactive social network analysis. SP will be designed on top of the classical Differential Privacy framework, and dedicated to overcoming the shortcomings of those anonymization-based solutions. Then, we will propose a “privacy-utility” quantification mechanism together with a noise selection algorithm to help strike the balance between the privacy and utility of the query results. Last but not least, an openly accessible, real data based social network analyzing platform will be built, with the hope that further studies will be inspired.

近年来，不断兴起的各种在线社交网络服务为我们带来了海量的社交网络数据。为了挖掘其中蕴藏的价值，人们展开了各种各样的社交网络分析。与此同时，作为真实社会的写照，社交网络数据包含了大量的个人信息。因此，在进行社交网络分析的同时，我们需要对数据中可能涉及的用户隐私进行保护。目前，国内外面向社交网络分析的隐私保护研究已经取得了不少成果，但总体来看仍处于起步阶段，还存在着许多局限和不足。在本项目中，我们将针对现有的基于匿名化发布的隐私保护方案的缺陷，从传统数据库领域的差分隐私保护模型出发，建立面向交互式社交网络分析的无缝隐私框架。然后，以之为基础，我们将提出交互式查询结果的“隐私度-可用性”量化模型，使得噪声选择算法能够在查询结果的隐私度和可用性之间取得平衡。最后，我们还将搭建一套开放式的、基于真实社交网络数据的隐私保护分析平台，以期未来能够进行更加深入的研究。

随着云计算和移动互联网等技术的飞速发展，各种各样丰富的在线应用与服务给人们的生活带来了巨大的便利。然而，受限于用户设备特别是移动设备的计算和存储能力，许多计算和存储任务都被托管到了云端，由此给用户数据的隐私保护带来了新的挑战。.本项目主要研究了三方面问题：.1、针对可信云端和图数据，在不对外发布数据、仅提供数据查询接口的前提下，如何对查询结果加扰以保护用户隐私；.2、针对不可信云端和图像数据，如何将用户的隐私数据从图像中分离并加密存储，同时还能支持用户在云端对加密后的隐私数据的搜索；.3、针对不可信云端和图像数据，如何将无关人员从图像中识别并消除，从而保护其肖像隐私。