Abstract: In order to predict the likelihood of network risks accurately and in real-time, and help the administrators to manage network risks effectively， a Time-Varying Markov Model (TVMM) for real-time risk probability prediction was proposed. A real-time risk probability prediction method, which is able to predict the probability of network risk in future exactly with a real-time-updating-state probability transition matrix of TVMM, was presented. Combined with the theory of feature extraction and statistical learning, the model was used to calculate the risk probability of the network at different risk level in network attack environment. The result shows that TVMM has higher real-time, objectivity and accuracy than the traditional Markov model.

Key words: computer application, time-varying Markov model, security risk prediction, network attack