Privacy Preserving Data Publishing in Social Network

PhD Thesis Proposal Defence


Title: "Privacy Preserving Data Publishing in Social Network"

by

Mr. Mingxuan Yuan


ABSTRACT:

Nowadays, more and more people join social networks, such as Facebook, 
Linkedin, and Livespace, to share their own information and at the same time to 
monitor or participate in different activities. It is a great opportunity to 
obtain useful information from these social network data. Meanwhile, the 
information stored in the social networks are under high risk of attack by 
various malicious users, in other words, people's privacy could be easily 
breached via some domain knowledge. Thus, for a service provider, such as 
Facebook and Linkedin, how to publish a privacy preserving graph becomes an 
important problem. It is essential to protect users' privacy and at the same 
time provide ``useful'' data. Targeting on the privacy preserving graph 
publication problem, in this thesis, we propose four graph publishing models 
and one collaborative protocol, which cover different aspects of graph 
publication issues.

1) We propose a novel privacy preserving graph construction technique based on 
adding noise nodes. This new graph construction algorithm provides privacy 
protection on the individuals in the graph and their attributes as well as 
maintaining good graph utility.

2) To increase the utilities of the published graph, we also propose a 
personalized protection framework for social networks, which publishes graphs 
with the consideration of users' personalize privacy setting.

3) After observing the lack of related work on the weighted graph model, which 
is general for online social network application, we propose graph protection 
models which protect individuals when the weights on the relationships are 
considered.

4) We also propose a new protection model for the attacks which are based on 
analyzing the label-structure relationship.

5) Due to the unavailability of the trusted centralized data owners across 
multiple social graphs, we propose a Secure Multi-Party Computation (SMC) 
protocol, which securely generates a privacy preserving graph in a distributed 
environment.

We conduct extensive experiments for the privacy preserving graph publishing 
models and the protocol above on various datasets. The experiments showed the 
efficiency and effectiveness of our solutions.


Date:                   Tuesday, 20 September 2011

Time:                   2:00pm - 4:00pm

Venue:                  Room 3401
                         lifts 17/18

Committee Members:      Dr. Lei Chen (Supervisor)
                         Prof. Cunsheng Ding (Chairperson)
 			Dr. Raymond Wong
 			Dr. Ke Yi


**** ALL are Welcome ****