A social network system has failure characteristics for random attacks of components or target attacks. This paper constructs related models for complex network defence systems to support the integrity of the social network system. First, we discuss the impact of component failure on complex systems and determine the risk scope. Second, based on the attack tolerance of the percolation theory, we verify the robustness of the network system through the percolation threshold fc to determine its optimal distribution. Third, we build a bimodal-distributed network model based on the network optimality to resist network failure. The model simulation results show that when the degree node is Kmin > 1 and Kmax = AN2/3 in the complex networks, these nodes themselves form a largest cluster to guarantee the integrity of the network system, and to ensure that the network is still robust to subsequent attacks after the removal of the central hub nodes.