P3: A Privacy Preserving Personalization Middleware for recommendation-based services
29 July 2011
We propose the design of a privacy-preserving-personalization middleware that enables the end-user to avail of personalized recommendations without disclosing sensitive profile information to the content/service-provider (or any third party for that matter). Our solution relies on a distributed infrastructure comprising of local clients running on end-user devices and a set of middleware nodes that could be collaboratively donated by few end-users or hosted by multiple noncolluding third parties. The key idea is to locally compute the user's profile on the device, locally determine the interest group of the user wherein an interest-group will comprise users with similar interest, and anonymously aggregate the collective behaviour of the members of the interest group at some middleware node to generate recommendations for the group members. In addition, our system is also open for third party content and recommendation injection without leaking the users privacy.
