Most research of network security based on the traditional game theory use completely rational assumption.For the condition of bounded rationality and the characteristic of dynamic changing process

we combined the evolutionary game model with Markov decision-making process based on the theory of non-cooperative evolutionary game

which is on the restraint of bounded rationality.Thus we constructed a multi-stage Markov attack-defense evolutionary game model to achieve multi-stage and multi-state dynamic analysis and evolution.Besides

on the basis of the sum discount payoffs to design objective function

the method to solve the evolutionary stable equilibrium was proposed

and the optimal defense strategy selection algorithm was provided.Finally

the validity of the model and method is validated by numerical simulation.