To evaluate the large-scale computer networks

a Bayesian-network-approximate-reasoning-based method for vulnerabilities evaluation was proposed.First

it models the elements which compose the network and the factors which affect the network security.Second

it builds the attack state graph（ASG） of the computer network to describe the process of vulnerability exploitation.Then

it makes the approximate reasoning to the ASG by stochastic sampling.At last

after the samples analysis and statistic

it achieves the quantitative evaluation result and will provide the theoretical evidence to improve the network security.