ARMOR: A trust-based privacy-preserving framework for decentralized friend recommendation in online social networks

Friend recommendation in online social networks (OSNs) has recently experienced rapid development and received much research attention. Existing recommender systems on the basis of the big social data mostly employ centralized framework, which would cause lots of problems, such as single point failure, communication bottleneck and so on. Some other studies focus on decentralized framework for recommendation, however, most of them concentrate on the improvement of recommendation quality, while underestimating privacy issues, e.g. OSN users' privacy concerns regarding their social relationships, social attributes, and recommendation profiles. In this paper, we propose a novel decentralized framework, namely ARMOR, which utilizes OSN users' social attributes and trust relationships to achieve the friend recommendation in a privacy-preserving manner. In ARMOR, we adopt a light-weight privacy-preserving protocol to aggregate the utilities of multi-hop trust chains and compute the recommender results securely. We also analyze the efficiency of ARMOR in theory and prove that OSN users' privacy can be preserved. Finally, we conduct an experiment to evaluate ARMOR over a real-world dataset and empirical results demonstrate that our ARMOR can effectively and efficiently recommend friends in a privacy-preserving way.