In social networks, many applications and spreading depend on the nodes with high influence to do viral marketing, which indicates that nodes' influence should be measured in a comprehensive and reasonable way. The appearance of fake fans results in change of network topology and brings new challenge to topology-based traditional methods. This paper incorporates both the network topology and interactions among nodes into our new distribution mechanism of node influence calculation in social networks. Considering the similarity of node behaviors in time domain and several key factors, this paper presents by a discounted Bayesian model for direct influence between nodes at first. Then a semi-ring-based aggregation implements for indirect influence and the composite influence are obtained by the combination of both direct and indirect influences. Simulation shows that this mechanism not only performs well against fake fans attack and restrains the fluctuation of nodes' influence, but also spreads to more nodes when we choose several nodes with high influence under our method to be source nodes.