Towards Privacy in a Context-Aware Social Network Based Recommendation System

Previous work has reported on `Instant Knowledge', a context-aware social networking based recommendation system for enterprise. This paper outlines a hierarchical privacy architecture, to provide anonymity, unlink ability, unobservability and pseudonymity to IK users. Users are grouped according to `proportional distance reservation', which indicates how likely users are willing to share private information. The protection of private information is stronger when queries are made by `distant' users, and weaker for fellow group members.