Through study on methods for forecasting multi-stage attack,we proposed a forecasting approach based on fuzzy,Hidden Markov Model(HMM) by integrating the association rule,fuzzy evaluation method and hidden Markov model together.Firstly,the original alarm information was fused into hyper alarm information.Secondly,the initial state matrix was obtained by the initial probability of the attack behaviors,the state transition matrix was determined according to the association rule,and the observation matrix was obtained by fuzzy evaluation.Finally,the attack scenarios leading to the alarm information were recognized by the Forward algorithm of HMM,and the next possible attack sequence was forecasted by the Viterbi algorithm of HMM.The results of simulation experiments verify the validity of this approach.