Aiming at the fast and accurate network fault detection and diagnosis, combining the danger theory with dynamic clonal selection algorithm, a danger theory based immune model used for network fault detection is proposed. And according to the characteristics of network failure to improve danger theory and dynamic clonal selection algorithm.Firstly, the concentration of antigen danger signal is recognized by using the danger theory model,and using the mature detector to detect known fault type. Then an improved clonal selection algorithm is set up to learn the unknown fault types. Experiments were undertaken with various type of network fault diagnosis to demonstrate that the model not only has better detection rate and adaptability but also can improve the efficiency and accuracy.