Research On The Decision-making Method Of Network Defense Strategy Based On Dynamic Game Theory

With the rapid development of information technology,the form of network security has been altered dramatically,and the means of cyber-attack are becoming more sophisticated.The essential key of how to deal with the increasing number of network attacks is how to prompt the capability of the network defense.The traditional “passive defense” technology,which mainly consists of intrusion monitoring,firewall and vulnerability scanning technology,has been unable to effectively deal with the current flexible network attacks.In view of above situation,the proactive defense technology of “changing the rules of the game” arises at the historical moment.In the network attack-defense scenario,the essence of network attack and defense lies in confrontation.As one of the basic theories of proactive defense technology,game theory can solve the decision-making problems in the confrontational environment,and is an effective guiding theory for the selection of network attack and defense strategies.Therefore,this paper researches the approaches of selecting strategies of network defense on the dynamic game theory.Main results are summarized as follows:1.Most of the existing network attack-defense studies focus on one side of game players,and there are some problems,such as it is difficult to depict and display the network attackdefense situation.There is no method to analyze the relationship among the network attack,defense behavior and the situation evolution trend effectively.Drawing on infectious disease dynamics theory,the paper constructs a network attack-defense simulation model based on game theory.On this basis,the experimental environment is built by using Net Logo simulation tool.In the end,the simulation results are summarized,which can support and guide the selection of network attack in different scenarios.2.Concerning about the characteristics of network attack and defense,some researchers put signal game theory into practice and study.However,in the process of income quantification,the phenomenon of unequal income between the two parties and the multi-stage confrontation process are ignored,so the models and methods are lack of universality.Aiming at this problem,the paper builds an active defense model based on non-zero-sum signal game theory,and puts forward a quantitative method according to the characteristics of network attack and defense.Then,the current optimal defense strategy algorithm is proposed based on the analysis of Nash equilibrium and refined Bayesian equilibrium.At last,experiments of the simulation show that the proposed model and method are effective and feasible,which can provide theory support for the selection of defense strategies in the procedure of discrete multi-stage attack-defense confrontation to some extent.3.The researchers apply stochastic game theory to network attack and defense modeling,but adopt completely rational assumption in the modeling process,ignoring the information opacity in the actual attack and defense process.In response to this problem,this paper constructs an attack-defense game model based on stochastic game theory under the constraint of bounded rationality,and adopts a state compression method based on service nodes.On this basis,a Defense-A3 C reinforcement learning model is proposed.Finally,experiments verify that the Defense-A3 C model can fit the best defense strategy after training.Compared with A3 C algorithm,Defense-A3 C has faster convergence speed and higher stability,and can provide some solutions for reinforcement learning to solve the problem of network defense strategy selection.