A Distributed Trust-based Recommendation System on Social Networks

In everyday life, we seek suggestions from people we know for deciding the best place to buy a particular good or service. In this work, we put forth a framework of an automated distributed recommendation system on a social network that exploits the widely studied concept of trust, to get personalized responses. The main contribution of our model is to combine two forms in which trust is perceived, the friendship trust and domain-expertise based trail levels, to efficiently propagate a query on a social network. We empirically validate the role of trust in online social networks by crawling the online social networking site Orkut and evaluate our recommendation system against one in which trust has not been used. We also analyze the results by varying different parameters of our model. Our framework is designed to work in a distributed environment and can thus be useful with upcoming technologies such as Mobile Ad Hoc networks, P2P networks and Semantic Web.