With considering different anti-attack abilities of nodes and nonuniform transmission of edges in real networks, an SIR model is proposed to investigate the threshold behaviors of epidemic in scale-free networks based on the mean-field theory. In the proposed model, a vulnerability function related to node degree is introduced to describe the diversity of nodes’ anti-attack abilities, and a transmission function related to node degree is also introduced to describe the nonlinear infectivity of nodes. Theoretical analyse and numerical simulations show that the diversity of node anti-attack ability and the nonuniform transmission of diseases on edges can yield a positive epidemic threshold,meaning that the risk of epidemic prevalence is reduced.