The idea of eco-industrial parks (EIPs) is a new industrial organization form that reconciles the three dimensions of sustainability: society, economy and environment. EIPs represent a practical way to apply the concept of industrial symbiosis (IS). Industrial Symbiosis Network (ISN) is a complex system consisting of a variety of enterprises and embedded relations. The keys of ISN are collaboration and synergistic possibilities offered by geographic proximity. Since the 1970s, the EIPs practice is conducted in many countries, i.e., Denmark, USA, Australia, Japan, and China and so on. However, many factors such as technological innovation, combination, takeovers and so on have a great negative influence on the practice of EIPs. Therefore, it is very important to make scientific analysis of the vulnerability of EIPs, thus providing effective measures to develop ISN. This paper employs the complex network theory to analyze the statistical properties of symbiotic network in EIPs. The nodes in EIPs represent the enterprises, and the edges between the nodes denote the symbiosis relationships. So the complex network theory is supposed to be effective in describing the real ecological industrial system. The complex characteristics of industrial symbiosis network are analyzed from three aspects: (1) The enterprises in EIPs form industrial ecological chains networks involving horizontal and vertical coupling; (2) The diversity of the enterprises made the node complexity; (3) The ISN is affected by a wide variety of factors. After describing the case of certain coal chemical eco-industrial park (abbreviated as CCEIP hereafter), we establish the topology of the EIP by employing the NetDraw tool of Ucinet (version 6.0) software. The statistical characteristics of the network in CCEIP are then examined. An index system including degree, degree distribution, clustering coefficient, average shortest distance length is developed to measure the topological properties. Results show that CCEIP's average shortest distance length (3.6759) and clustering coefficient (0.1235) are both higher than those of random network (2.6780 and 0.0686 respectively). The degree follows power-law distribution. Therefore, we can infer that CCEIP has the small world and scale-free characteristics. As the security and stability of the nodes with maximum load greatly affects the network vulnerability, the paper adopted a strategy of deliberately attacking the nodes with maximum load, such as the anthropogenic bankruptcy of enterprise. Two important indexes, the network efficiency and the largest connected component, were used to analyze the network vulnerability and to measure the damage degree of cascading failure. The network efficiency declines tendency when the nodes of larger load are attacked. When the tolerance parameter is in the range of 0-0.3, the network efficiency and the largest connected component change abruptly.