Telecommunications network is vulnerable to large-scale infrastructure failures caused by physical attacks or natural disasters. Academics view this failure as region failure, which is geographically related and highly localized. Several simplified models have been previously proposed, resulting in an inaccurate characterization of network activity. In this paper, we propose a probability model with full consideration of the actual situation, which gives a more realistic picture of network behavior. In particular, the disaster takes the form of a randomly placed disk in a plane. The research aims to find out disaster locations that can cause significant damage to network performance, i.e., vulnerable areas of the network. The topology of vulnerable areas can be redesigned to improve the network’s disaster survivability. Moreover, only a few special cases need to be judged for the sake of achieving the aim despite disasters occur randomly across the plane. Another primary contribution is our attempt to construct a proper network topology by using clusters as units. Finally, numerical experiments conducted on this network topology demonstrate the applicability of the methodology in realistic scenarios. The work in this paper provides guidance for real-world problems and offers a way to figure them.