Trust propagation algorithm based on learning automata for inferring local trust in online social networks

Online social networks have provided an appropriate infrastructure for users to interact with one another and share information. Since trust is one of the most important factors in forming social interactions, it is necessary in these networks to evaluate trust from one user to another indirectly connected user, using propagating trust along reliable trust paths between the two users. The quality of trust inference based on trust propagation is affected by the length of trust paths and also different aggregation strategies for combining trust values derived from multiple paths. While evaluating trust value based on all paths provides more accurate trust inference results, it is very time consuming to be acceptable in large social networks. Therefore, discovering reliable trust paths is always challenging in these networks. Another important challenge is how to aggregate trust values of multiple paths. In this paper, we first propose a new aggregation strategy on the basis of the standard collaborative filtering. We then present a heuristic algorithm based on learning automata, called DLATrust, for discovering reliable paths between two users and inferring the value of trust using the proposed aggregation strategy. The experimental results conducted on the online social network dataset of Advogato demonstrate that DLATrust can efficiently identify reliable trust paths and predict trust with a high accuracy.