Abstract: High availability network usually are a huge number of malicious invasion of current early warning method can only analyze the history and the current network intrusion status, cannot predict the future high availability network invasion situation, in order to improve the accuracy of early warning, put forward a new high availability network intrusion early warning method, this paper introduces the hidden markov model, on this basis, the building based on the HMM model of real-time high availability network intrusion early warning model, according to the analysis of risk theory to the evaluation result, the high availability network state was divided into four grades, by building the HMM-NSSP model to realize the network intrusion early warning. In order to improve the early warning of the HMM-NSSP model precision, the introduction of the maximum entropy method for high availability network security situational consistency judgement, to complete the high availability of network intrusion early warning method improvement, the experimental results show that the proposed method the early warning results and actual results are basically identical, has the very high accuracy of early warning, a high availability.