Research on network robustness based on different deliberate attack methods

• Published in: Physica A 2020-05, Vol.545, p.123588, Article 123588
• Main Authors: Yang, Guizhen, Qi, Xiaogang, Liu, Lifang
• Format: Article
• Language: English
• Subjects: Cascading failure; Critical threshold; Deliberate attack; Robustness
• ISSN: 0378-4371; 1873-2119
• DOI: 10.1016/j.physa.2019.123588

Many real-world networks can be abstracted into complex networks, which nodes are correlated with each other, once these functional nodes suffer from external or internal attack, they may lead to malfunction of the rest part of networks. In order to explore the impact of different attack methods on network robustness against cascading failures, we propose three attack methods and compare them with HD and LD which had been researched. By simulation on ER, BA and WS networks, the results show that, no matter what value α is, for BA and WS networks, attacking networks in GHS is the most difficult to cause cascading failures, and now, the networks show the strongest robustness. In α1, it is the easiest to make networks cascading failures when attacked networks by HD. In α=1, attacking networks by MD and LD,respectively, is the easiest to lead networks to cascading failures and the networks show the worst robustness. However, for ER network, no matter what value α is, there is the same effectiveness of attacking network by five attack methods and has stronger robustness than BA and WS networks. This work provides a good reference value for maintaining network security and enabling the network to operate normally and stably. It is possible to make effective preventive measures against these different attacks in advance and save costs for maintaining the network. •Propose three attack methods and compare them with HD and LD.•No matter what value α is, for BA and WS networks, the robustness against cascading failures is the best by GHS.•No matter what value α is, For ER network, there is the same effectiveness of attacking network by five attack methods and has stronger robustness than BA and WS networks.