With the wide application of computer network, network security has attracted more and more attention. The main reason why all kinds of attacks on the network can pose a great threat to the network security is the vulnerability of the computer network system itself. Introducing neural network technology into computer network vulnerability assessment can give full play to the advantages of neural network in network vulnerability assessment. The purpose of this article is by organizing feature map neural network and the combination of multilayer feedforward neural network, the training samples using SOM neural network clustering, the result of clustering are added to the original training samples and set a certain weight, based on the weighted iterative update ceaselessly, in order to improve the convergence speed of BP neural network. On the BP neural network algorithm for LM algorithm was improved, the large matrix inversion in the LM algorithm using the parallel algorithm method is improved for solving system of linear equations, and use of computer network vulnerability assessment as the computer simulation and analysis on the actual example, design a set of computer network vulnerability assessment scheme, finally the vulnerability is lower than 0.75, which is beneficial to research on related theory and application to provide the reference and help.