Research On Intelligent Disintegration And Deception Defense Method For Heterogeneous Complex Network

The rapid development of intelligent technology is driving the evolution of warfare.The “mosaic” warfare is a new and vital operational concept proposed by the U.S.military in recent years,showing intelligent operational characteristics and leading the development of the U.S.military intelligence,whose main idea is to build a dynamic reorganization of the “kill network” in an adaptive manner.Using scientific theory to model the“kill network accurately”,the use of intelligent technology to achieve assisted planning and decision-making is an important research issue to be resolved.On the one hand,the attacker needs to quickly find the critical nodes in the network from the enemy’s complex kill network to weaken the enemy’s combat capability.On the other hand,the defender needs to take advantage of the asymmetric information in the battlefield to create complexity for the active defense to achieve “Mobilize the enemy without being mobilized by the enemy.” To address the offensive and defensive problems mentioned above,we use complex networks,deep learning,and other theoretical approaches to conduct a systematic and in-depth study on three aspects of heterogeneous network modeling,network disintegration,and network defense.The main work and innovation points of the paper are as follows:(1)A heterogeneous kill network model for “mosaic” warfare is established,and heterogeneous kill network operational capability evaluation indexes are proposed.Firstly,the concept of heterogeneous network is introduced and further elaborated based on the concept of heterogeneous network.For different types of combat entities and the interaction between different combat entities,the heterogeneous network theory is used to model the military network and construct a heterogeneous kill network model.Second,for the characteristics of dynamic changes in the battlefield environment,the network structure information is integrated into the heterogeneous network kill chain,and the dynamic combat capability index is proposed to assess the combat capability of the heterogeneous kill network.The experimental results show that the dynamic combat capability index can reasonably and accurately evaluate the combat capability level of heterogeneous combat networks,and compared with the traditional gaint connection component size index,it can not only evaluate the network structure but also effectively assess the performance of different types of nodes constituting the kill chain in heterogeneous networks,and more comprehensively and accurately reflect the changes in the combat capability of the kill network.(2)An intelligence disintegration method for heterogeneous complex networks based on graph embedding learning is proposed.Firstly,a heterogeneous kill network disintegration strategy model is established.Secondly,a Heterogeneous Kill Network Disintegration Method Based on Graph Embedding via DQN(HDGED)is proposed.The HDGED method is innovatively designed to embed the node type information and structure information in the heterogeneous killing network into the representation vector by subtype encoding-aggregation,and then the representation vector decoding is trained to obtain the optimal tiling strategy by the deep reinforcement learning method.Finally,comparative experiments are done on heterogeneous kill networks.The results show that the proposed HDGED method can fully exploit the node type information and structural information of heterogeneous networks,quickly find the critical nodes affecting the network function,and better migrate for combat networks of different sizes.Its disintegration effect is improved by 46% compared with the baseline method.(3)A defense strategy for heterogeneous kill networks is proposed based on hiding deception under asymmetric information conditions.Firstly,by hiding the critical nodes in the network and inducing the attacker to attack the non-important nodes,a heterogeneous network defense strategy based on hiding deception under asymmetric information conditions is proposed,i.e.,the adversarial hiding deception strategy,which actively constructs the adversarial hiding deception network and overcomes the defects of the traditional passive defense strategy.Secondly,the integrated graph structure information and type information of nodes are integrated to design the node importance index and fitness function for heterogeneous kill networks and propose an Adversarial Hiding Deception Network optimization method based on Genetic algorithm(AHDNG).The method can optimally generate the adversarial hiding deception network by setting the hidden and deceptive edges of the network according to the constraints,and the generated network has a similar graph structure as the original network.The experimental results show that the combat capability of the original undefended network is wholly disrupted when the attack strength is 0.25 using the maximum degree attack method.In contrast,the combat capability of the network defended by the adversarial hiding deception strategy is still maintained at 0.56.(4)An empirical study of offense and defense for typical scenarios of heterogeneous networks.Two typical scenarios of heterogeneous kill networks and infrastructure networks are empirically studied for offense and defense.Firstly,a heterogeneous kill network model is established based on operational assumptions,and secondly,two disruption strategies are used to disrupt them.The experimental results show that the HDGED method can disrupt the kill network under the attack strength of 0.16,while the baseline algorithm still has a remaining combat capability of 0.6 after the disruption.Further,the integrated verification of the attack and defense methods is carried out,and the defender uses the AHDNG method to defend the original network with hiding deception.The experimental results show that the combat capability of the kill network can still be maintained at 0.84,which provides auxiliary decision support for the attack and defense confrontation on the combat network.Second,the heterogeneous network functional chain is instantiated as a grid functional chain,and the functional chain is applied for the first time in the grid context to assess the power network capability.The experimental results show that the HDGED method can disintegrate the power network under the attack strength of 0.23,while the baseline algorithm still has a remaining grid capability of 0.38 after disintegration.The defender uses the AHDNG method to defend the original network against hiding deception,and the experimental results show that the grid capacity can still be maintained at 0.78.The two experiments verify the effectiveness and practicality of the proposed method and show that the proposed model and method can be extended to more networked and systematic complex system attacks and defenses.