Research On Attack Strategies For Complex Networks

With the development of science and technology, the study of complex network has become public. At the same time, the security of the complex network becomes significant than ever before due to various attacks. Usually, attacks include random attacks and selective attacks. The study of the invulnerability of complex networks to different attacks is important.Particularly, the attack strategy is crucial to the invulnerability of complex networks.In this paper, we firstly analyze the invulnerability of various complex network models and the attack efficiency of classical strategies based on degree centrality and betweenness centrality. Then, we propose two kinds of new attack strategies, namely, joint attack strategies and entropy based attack strategies. The former kind attacks include IDB(Initial degree and betweenness) and RDB(Recalculated degree and betweenness), the latter ones consist of IE(Information entropy), CE(Cross entropy) and LE(Local entropy). Lastly, we measure the attack efficiency of the new proposed strategies.The main contribution of this work includes:(1) The construction and analyze of complex network models. In order to comprehensively study the invulnerability of various complex network models, we select three standard network models(ER random networks, WS small-world networks and BA scale-free networks) and three kinds of real-world networks, including the power-grid network, the Coauthor network, and the Protein network.(2) The invulnerability of complex networks. The exponential networks have very strong robustness to random attack, the networks collapse when at least 80% of the nodes are removed. However, at least 40% of the nodes need to be removed even at the most effective attack(RBC strategy). The BA scale-free network behaves vulnerable. It collapses when 40%of the nodes are removed under static attacks. It only needs to remove 20% of the nodes under dynamic attacks. The three kinds of real-world networks are very fragile under both random attack and selective attacks, the topology structure of the network will be largely damaged when a small number of hub nodes are removed.(3) The analysis of the attack efficiency of the new proposed strategies. We firstly analyze the attack efficiency of classical attack strategies, and then we measure the new proposed attack strategies and compared with the classical attack strategies. To the WS small-world network, the attack efficiency of IDB strategy is improved by 20% to IDC strategy, and RDB strategy is improved by 30% to RDC strategy. To the power-grid network, the IDB strategy isthe best one, whose attack efficiency is much higher than other four strategies. For the ER random network and the BA scale-free network, the attack efficiency of ICE strategy is the best, and ICE strategy is improved by 25% to IBC strategy in the BA scale-free networks.In summary, the new proposed strategies of IDB strategy, RDB strategy, and entropy based attack strategy behave a better advantage over the classical attack strategies on various complex network models and real-world networks.