The current research of network security based on game theory fails to analyze the real-time

continuous

and random network attack and defense process.For the randomness of security states and the real-time character of network defense decision-making

we analyzed the network attack and defense behaviors from the view of dynamic and real-time confrontation.Then we combined and extended the differential game model and Markov decision-making method.On these basis

a Markov attack-defense differential game model is constructed

which can be adopted to analyze the multi-stage attack and defense process with short duration in each stage.Besides

a multi-stage game equilibrium solution is proposed

and an optimal defense strategy selection algorithm is designed.Finally

the experiments demonstrate that the model and method proposed in this paper are valid.