Vulnerability Of Complex Networks Under New Attack Strategies

In recent decades,many researchers,represented by Barabasi and Holme,have analyzed various complex model networks and real networks,according to random or targeted attack strategies,and have found many constructive conclusions.These attack strategies do not require the nodes and edges to be adjacent as they are removed.However,by analyzing the new attack mode in recent years,the analyzed results indicate that the nodes destroyed by the attackers are often adjacent.In addition,with the wide spread of the"robust and fragile",the node of max degree/betweeness has been well protected.It is difficult to be attacked/destroyed directly.Thus,the neighbor nodes of max degree/betweeness node are generally chosen as the first node of attack.Because the prevention of these neighbor nodes is relatively weak to be destroyed,they are the preferred targets by attacker.It is distinctly different from previous random and targeted attack strategies.Based on the above analysis,four types of new attack strategies:tree attack strategies,approximate longest path attack strategies,the shortest path attack strategies,neighbor nodes attack strategies,are adopted in this dissertation.The vulnerability of some complex networks is analyzed including three model networks?Random,Scale-free and Small-world?as well as eight real networks?USAir,NetSience,CRH,et al?.To sum up,the specific research works of this dissertation is as follows:?1?Vulnerability of complex networks under tree attack strategies.According to the attacker continues to invade all adjacent nodes of the node that have been invaded,and will destroy all invaded nodes at the same time.The four tree attack strategies are proposed,which are RRN,MaxDRN,MinDRN and Max BRN.The experiment results of model networks?Random,Scale-free and Small-world?demonstrate that Small-world network has the best robustness,while Scale-free network has the weakest robustness,and the robustness of Random network is between them,from LCC'vs.RN and L'vs.RN.Specifically,Max DRN and MaxBRN are the most destructive in the four tree attack strategies from LCC'vs.Step and L'vs.Step,especially for Scale-free network.Furthermore,under any one of the four tree attack strategies,both Random network and Small-world network shows that the larger the<k>,the smaller the Step.In addition,the experiment results of six real networks?Word adjacencies,USAir,NetSience,Polblogs,Football and Roget?have verified the conclusion that MaxDRN and Max BRN have the strongest damage to network connectivity.?2?Vulnerability of complex networks under approximate longest path strategies.Based on Pu‘s RPA approximate longest path attack strategy,two approximate longest path attack strategies of BLP and DLP are proposed.The experiment results of model networks?Random,Scale-free and Small-world?demonstrate that Small-world network has the best robustness,while Scale-free network has the weakest robustness,and the robustness of Random network is between them,from LCC'vs.RN and L'vs.RN.Specifically,BLP and DLP have more damage than RPA.Furthermore,under any one of the three approximate longest path attack strategies,the Random network and Small-world network also show that the larger the<k>,the smaller the Step.In addition,the experiment results of two real networks?China Railway High-speed network and Beijing Subway network?have verified the conclusion that BLP and DLP have the stronger damage than RPA to network connectivity.?3?Vulnerability of complex networks under shortest path attack strategies.From the conclusion that?people prefer to choose the shortest path between two nodes in the network??the extended theory of Milgram‘s?six degrees of separation??,RSP,BSP and DSP,three shortest path attack strategies,are proposed.The experiment results of three model networks?Random,Scale-free and Small-world?demonstrate that Small-world network has the best robustness,while Scale-free network has the weakest robustness,and the robustness of Random network is in the middle,from LCC'vs.RN and L'vs.RN.Specifically,when the network is completely disconnected,RN_BSP is the lowest in three shortest path attack strategies.Furthermore,three model networks all show that the larger the<k>,the larger the Step.In addition,the experiment results of two real networks?China Railway High-speed network and Beijing Subway network?have verified the conclusion that RN_BSP is the lowest in three shortest path attack strategies.?4?Vulnerability of complex networks under neighbor nodes attack strategies.According to the neighbor nodes of max degree or max betweeness node are often chosen as the first node of destruction,two neighbor nodes attack strategies of NND and NNB are proposed.The experiment results of three model networks?Random,Scale-free and Small-world?demonstrate that Small-world network has the best robustness,while Scale-free network has the weakest robustness,and the robustness of Random network is between them,from LCC'vs.RN and L'vs.RN.Specifically,when the network is completely disconnected,three model networks show that Step_NND>Step_NNB and RN_NND>RN_NNB.Furthermore,the Small-world network and Random network also show that the larger the<k>,the smaller the Step.In addition,the experiment results of six real networks?Word adjacencies,USAir,NetSience,Polblogs,Football and Roget?have verified the conclusion that Step_NND>Step_NNB and RN_NND>RN_NNB;?5?The research results of vulnerability of three model networks are concluded and compared,under tree attack strategies,the approximate longest path attack strategies,the shortest path attack strategies,as well as neighbor nodes attack strategies.In addition,the same features and the different features of three model networks were analyzed.