A Private and Reliable Recommendation System for Social Networks

With the proliferation of internet-based social networks into our lives, new mechanisms to control the release and use of personal data are required. As a step toward this goal, we develop a recommendation system which protects the privacy of user answers while allowing them to learn an aggregate weighted average of ratings. Due to the use of social network connections, the querier obtains a more relevant and trustworthy result than what generic anonymous recommendation systems can provide, while at the same time preserving user privacy. We also give experimental performance results for our solution and several recently developed secure computation techniques, which is of independent interest.