Online social networks (OSN) are among the most popular web-based services. As the nature of OSN requires sharing of information, users are encouraged to upload their data to OSN in order to build up their profiles, which in turn enhances the functionality of OSN itself. Many OSN applications use their users' data in information retrieval. Though they claim that their users' privacy is well preserved, many of them suffer from privacy exposure. Recent studies of inference attack showed that users' descriptive attributes on OSN applications can be inferred through the applications' ranked retrieval interfaces, even though the attributes are marked as ``private". The reason is that some ranked retrieval interfaces take into account private attribute values in their ranking functions in order to provide more accurate results. As a result, adversaries are able to gain extra information about private attributes through query interfaces of social networks.
Compared with existing privacy preserving data publishing and data mining techniques, the challenge of privacy preservation of social network data behind Ranked Retrieval model lies in the fact that instead of generating aggregate results, Ranked Retrieval model outputs numerical ranks based on relative closeness coefficients of data points and queries. In particular, the challenge is modeling and providing the privacy guarantee while minimizing utility loss with respect to ranked results.
In this dissertation we consider the privacy leakage through Ranked Retrieval model. We begin this dissertation by identifying two classes of adversaries based on their knowledge of the distribution of victim private attributes. In the first part of this dissertation, we present Equivalent Set, a privacy-preserving framework that involves the design of ranking functions. In the second part of the dissertation, we present Polymorphic Value Set. In addition to the design of ranking functions, Polymorphic Value Set enables polymorphism of private attributes by allowing them to respond to different queries in different ways. We consider utility loss of ranked results and develop implementations for different classes of adversaries. We valid our implementations through theoretical analysis and extensive experiments over real-world social network datasets.
|Advisor:||Cheng, Xiuzhen, Zhang, Nan|
|Commitee:||Choi, Hyeong-Ah, Huang, Hao Howie|
|School:||The George Washington University|
|School Location:||United States -- District of Columbia|
|Source:||DAI-B 81/2(E), Dissertation Abstracts International|
|Keywords:||Database management, Privacy and security, Rank inference, Social network|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be