Thesis Topic Details

Topic ID:
The potential of Recommender Systems to breach user privacy in P2P.
Guillaume Jourjon
Research Area:
Privacy, Peer-to-Peer
Associated Staff
Salil Kanhere
Topic Details
R & D
Group Suitable:
Recommender systems are a class of personalized systems that recommend to their users the items they may wish to examine or consume. Recommender systems research produced a variety of methods deployed in numerous applications and Websites.
In this project, we focus on the potential of recommendation systems from a user privacy perspective. First, we aim to design a new model based on users interactions with P2P file sharing systems. Two characteristics are of interest as for the consumption model. First, downloaded content is presumably free in P2P, and hence items prices are not taken into account. Second, the decentralized nature of P2P allows relaxing the hypothesis that users preferences are not biased by the recommendations selected by the system itself. P2P users deliberately download content from the P2P system with their choices directed only by social, geopolitical or cultural choices. It would then be natural to observe higher similarity between users and apply different record linkage techniques, inherited from Recommendation systems to predict users download. We will then dig into the capabilities of de-anonymizing users downloads even though they are hidden (e.g. ToR, proxies). Ideally, we should get with some insights on how users can avoid this linkage.
The student will work closely with researchers in the NICTA Networks Research Group. The project will be primarily based at the NICTA ATP laboratory, where the work environment is a group comprised of junior and senior research staff, and PhD students. IT support and resources will be provided by NICTA.
Past Student Reports
No Reports Available. Contact the supervisor for more information.

Check out all available reports in the CSE Thesis Report Library.

NOTE: only current CSE students can login to view and select reports to download.