Research On Network Attack And Defense Competition Based On Differential Game

With the rapid increase of the wireless network scale and the extensive interconnection between wireless terminals,the boundary of wireless network is becoming blurred and the security challenges are more and more severe.The increasingly complex and diverse forms of network attacks,such as alliance interaction,resource competition,risk propagation,put forward higher requirements for the traditional defense mechanisms in the current wireless network system.It is of great significance to grasp the basic theory of network security,analyze the characteristics of network attacks and combine the relationship between attackers and defenders to formulate defense strategies.Facing large-scale pervasive-connection wireless networks,this thesis carries out the study of basic game model on network attack and defense competition based on differential game,which can provide theoretical guidance for dynamic network attack-defense modeling and real-time defense strategy selection.The specific research results are as follows:(1)Considering the alliance interaction characteristics of security attack-defense in the future network,a multi-attacker-multi-defender model based on differential game is proposed to analyze the interaction between attack alliance composed of multiple attack nodes and defense alliance composed of multiple defense nodes.The number of attack/defense nodes is used to model the attack/defense alliance strength.The attack/defense alliance adjusts the alliance strength by waking up the sleeping nodes or retreating the attack/defense nodes,resulting in the transition of weak alliance nodes to paralyzed nodes.In order to reduce the ping-pong effect in attack and defense paralysis,paralysis thresholds are introduced to restrict the paralysis results.The evolution of attack/defense alliance strength is described by constructing the state transition differential equations.On this basis,a differential game model is established to analyze the continuous real-time game states between attack and defense alliances.The game equilibrium strategy is solved based on Hamilton optimal control method,and a multi-attacker-multi-defender optimal strategy selection algorithm is proposed to realize the optimal attack and defense strategy under game equilibrium.Simulation results show the evolution of the attack/defense alliance strength and the optimal attack/defense strategies.The results show that attack/defense alliance tends to enhance the strength at the beginning of the game,and then gradually weakens the strength to obtain the ideal cost-benefit ratio.In addition,the defense effectiveness can be improved by increasing the defense paralysis threshold.(2)Considering the competition for network devices between the attacker and defender,an attack-defense resource competition model based on differential game is proposed,where the network devices are regarded as the competed objective,i.e.,network resource,between the attacker and defender.The attacker/defender adjusts the competition strength injected into devices to get control of these devices,resulting in the state transition of the devices.By analyzing the state transition process of each device,differential equations are established to describe the evolution of the overall network security state.Based on the construction of zero-sum differential game model,the optimization problems of both game players are solved based on optimal control theory,and an optimal strategy selection method based on Gauss-Seidel like implicit finite-difference method is proposed to obtain the saddle point strategy.Simulation results show the evolution of network security states in the attack-defense resource competition,and indicate that the proposed game model is more effective than the non-game scheme in improving system security.(3)Considering the attack-defense propagations caused by the strong connection of close neighbors among massive nodes in large-scale pervasive-connected networks,a network attack-defense resource competition model for propagation behavior is proposed to model and analyze the network resource competition and attack-defense strategic interaction.The security state of sleeping devices in the network is not only directly affected by the injected attack-defense competition strength,but also indirectly infected by the attack-defense propagations of neighbor devices.By modeling the attack and defense competition strength as game strategies,an attack and defense resource competition model based on differential game is established.On this basis,Hamilton optimal control method is used to optimize the game objectives,analyze the saddle point strategy and realize the optimal strategy selection.Results show the evolution process of network device security state under the influence of propagation behavior,and demonstrate that compared with the attackdefense propagation model and centralized control model,the proposed model can reduce the system security loss and present a better system security state under the attack-defense confrontation equilibrium.To sum up,considering three attack and defense problems,i.e.,alliance interaction,resource competition and propagation behavior in wireless network security,this thesis establishes an attack and defense model with the characteristics of time continuity,competitive dynamicity and strategy real-time based on differential game,which provides mathematical rationale and theoretical guidance for the prediction of network attack and defense behaviors and the design of effective security defense mechanisms.