Abstract: In social networking,trust calculation have attracted more and more attention,especially between two unknown nodes．At present most of the trust models aren’t accurate enough due to incomplete recommendation evidence．With the increasing of community,community becomes a development trend of social networks today.In this paper,we first proposed community recommendation model in substitution for the existing node recommendation model,to improve integrity and reliability of the recommend evidence,and consequently improve the accuracy of unknown nodes trust computation．Secondly friend group trust influence was considered and community correlation factor was presented to solve the community possible collusion attack．Finally,simulation results verified the rationality and effectiveness of the proposed method.

Key words: Social network,Community recommendation,Community correlation factor,Trust,Friend group