Research On Reconstruction And Disruption Of Complex Network And Its Application

Unmanned cluster operation is an important form of intelligent operations in the future.Recent technological advances have made the threat of this new type of operations more and more urgent.How to model,analyze,evaluate and plan the large-scale unmanned cluster system(UCS)and its confrontation is the key problem to be solved urgently in the future intelligent war.On the one hand,cluster confrontation accompanied by a large number of heterogeneous agents of the group game,is an extremely complex dynamic process,which is difficult to model and evaluate.Meanwhile complex network theory is an effective tool to describe and analyze complex systems,which provides a theoretical basis for the attack and defense modeling and evaluation of unmanned cluster systems.On the other hand,deceiving decision system of counterparty is an important means of intelligent confrontation between clusters.It makes the battlefield environment flooded with a large amount of incomplete information and false information,which seriously interferes with the effectiveness of the cluster’s offense and defense.The theory of complex network reconstruction is an effective way to restore the real network information,which provides a novel idea for uncovering the "fog of war" in the confrontation of unmanned cluster system.Therefore,this dissertation takes confrontation of unmanned cluster system as the background.There are two main concerns in this dissertation: one is complex network reconstruction under incomplete and spurious information;another is collapse and defense of complex networks under this information noise.Based on complex network theory,the problem analysis,system modeling,algorithm designing,and solution verification are studied.The main contributions are as follows:(1)By analyzing the limitations of the network reconstruction method under incomplete and spurious information,this dissertation focuses on two typical network reconstruction sub-problems(with prior knowledge or without prior knowledge)and proposes two complementary network reconstruction approaches:enhanced stochastic block models and similarity-based network reconstruction method.On the one hand,network reconstruction without prior knowledge is a fundamental theoretical problem.Although SBM(stochastic block models)method is a feasible way to solve this problem,the disadvantages of time-consuming and being trapped in a local optimal solution seriously reduce its practicability.This thesis defines and constructs the model of network reconstruction without prior knowledge and proposes a new approach by improving the solution speed and optimality of the SBM:enhanced stochastic block models(ESBM).On the other hand,compared with the former,network reconstruction with prior knowledge in the real world is a more complicated problem,and there is no clear definition and specialized solution in the field of complex networks.Taking node degree as the example of prior knowledge,this thesis presents a new problem: network reconstruction with partial node degree information.Considering both the micro and macro characteristics of the network,a similarity-based network reconstruction(network reconstruction based on prior knowledge,PKNR)method is proposed for the first time.The experimental results of the simulated network and the real network show that: first,the ESBM’s prediction accuracy is better than the SBM method without prior knowledge,and the convergence speed can be more than 10 times faster than the SBM;second,with prior knowledge,the reconstruction network generated by the PKNR method can reduce the relative error rate by 15.6%.Third,with prior knowledge,the prediction accuracy of PKNR is higher than the reconstruction method without prior knowledge,and the convergence speed is more than 100 times faster than ESBM,which indicates that prior knowledge and similarity index method are helpful to improve the efficiency of network reconstruction.(2)Focusing on the problem of attack and defense of complex network under incomplete and spurious information,this dissertation creatively incorporates network reconstruction capabilities and observation information camouflage capabilities,and proposes two novel methods for network attack and network defense: network disruption based on network reconstruction(NR-ND)and complex network defense method based on concealment and deception;This thesis reveals the "sharpening effect" of network reconstruction on network disruption and the characteristics of attack migration of hidden and deception strategies.For the attacker,the traditional network disruption method(network disruption based on network observation,NO-ND)is difficult to apply to the situation of incomplete and spurious information.This dissertation gives a novel idea: perform missing information prediction and spurious information identification before network disintegration.Based on this,this thesis creatively gives a novel network disruption method: network disruption based on network reconstruction(NR-ND).For the defender,the traditional defense methods(such as network rewiring)not only costs a lot,but also ignore the influence of incomplete and spurious information.Therefore,by analyzing the limitations of passive defense,this dissertation constructs an attack migration model based on the observation information camouflage.Based on the relationship between the node degree sequence and the camouflage of observation information,this thesis creatively proposes two complex network defense methods: observation network generation algorithm based on concealment and deception under specified degree sequence;observation network generation algorithm based on concealment and deception under limited resources.The experimental results of the simulated network and the real network show that: first,the disintegration effect of NR-ND is better than NO-ND,indicating that network reconstruction helps the network disruption;second,the disruption effect of NR-ND is even superior to that achieved by complete information when the noise is relatively small,and the thesis names this phenomenon as "sharpening effect" of network reconstruction;third,with the assistance of concealment and deception,the robustness of the defender’s network structure is increased by up to44.17%,which indicates that network defense based on concealment and deception can effectively migrate attack damage and improve the overall defense capability of the network.(3)Aiming at the problem of confrontation of unmanned cluster system under incomplete and spurious information,from the perspective of network science,this dissertation introduces the theory of complex network reconstruction,collapse and defense into the UCS for the first time and systematically puts forward a set of analysis methods,which provides a new way of thinking for the research of attack and defense of unmanned cluster system.Unmanned clusters have the characteristics of emergence mechanism and self-organization,which are very similar and isomorphic to complex networks.However,little contribution is based on complex network theory under incomplete and spurious information.Therefore,based on the complex network model,network topology structure and dynamic reconfiguration mechanism of unmanned cluster system,this dissertation introduces the complex network theoretical achievements of this thesis into the unmanned cluster system,and proposes attack and defense method for the UCS.Case experiments show that this method can effectively characterize the confrontation between unmanned cluster systems,effectively improve the offensive and defensive capabilities of the UCS,explore the offensive and defensive problems of the UCS from a new perspective,and provide key technical support for unmanned cluster operations.