Enhancing Privacy in Online Social Communities: Can Trust Help Mitigate Privacy Risks?

Abstract

The context based privacy model (CBPM) has proved to be successful in strengthening privacy specifications in social media. It allows users to define their own contexts and specify fine-grained policies. Collective-CBPM learns the user policies from community. Our experiments on a sample collection of Facebook data demonstrated the models feasibility in real time systems. These experiments however, did not capture all of the user scenarios; in this paper we simulate users for all possible user scenarios in a social network. We operationalize the C-CBPM model and study its functional behavior. We conduct experiments on a simulated environment. Our results demonstrate that even the most conservative user never incurs risk greater than 20%. Moreover, the risk diminishes to 0 as the trust increases between donors and adopters. The model poses absolutely no risk to other liberal or semi-liberal users.