κ-FuzzyTrust: Efficient trust computation for large-scale mobile social networks using a fuzzy implicit social graph

Large-scale mobile social networks (MSNs) facilitate connections between mobile devices and provide an effective mobile computing environment in which users can access, share, and distribute information. In MSNs, users may belong to more than one community or cluster, and overlapping users may play a special role in complex MSNs. For such MSNs, a key problem is how to evaluate or explain user trustworthiness. In this context, trust inference plays a critical role in establishing trusted social links between mobile users. To infer fuzzy trust relations between users in MSNs with overlapping communities, we propose an efficient trust inference mechanism based on fuzzy communities, which we call κ-FuzzyTrust. We propose an algorithm for detection of community structure in complex networks under fuzzy degree κ and construct a fuzzy implicit social graph. We then construct a mobile social context including static attributes (such as user profile and prestige) and dynamic behavioural characteristics(such as user interaction partners, interaction familiarity, communication location and time) based on the fuzzy implicit social graph. We infer the trust value between two mobile users using this mobile social context. We discuss the aggregation and propagation of trust values for overlapping users and indirect connected users. Finally, we evaluate the performance of κ-FuzzyTrust in simulations. The results show the validity of our fuzzy inference mechanism for behavioural trust relationships in MSNs. They also demonstrate that κ-FuzzyTrust can infer trust values with high precision.