We study the resilience of complex networks against attacks in which nodes are targeted intelligently, but where disabling a node has a cost to the attacker which depends on its degree. Attackers have to meet these costs with limited resources, which constrains their actions. A network's integrity is quantified in terms of the efficacy of the process that it supports. We calculate how the optimal attack strategy and the most attack-resistant network degree statistics depend on the node removal cost function and the attack resources. The resilience of networks against intelligent attacks is found to depend strongly on the node removal cost function faced by the attacker. In particular, if node removal costs increase sufficiently fast with the node degree, power law networks are found to be more resilient than Poissonian ones, even against optimized intelligent attacks.
Advancing network resilience theories with symbolized reinforcement learning
Beijing National Research Center for Information Science and Technology, Tsinghua University, Beijing, China(北京信息科学与技术国家研究中心,清华大学,北京,中国);Department of Electronic Engineering, Tsinghua University, Beijing, China(清华大学电子工程系,北京,中国);Department of Computer Science, Rensselaer Polytechnic Institute, Troy, NY, USA(罗切斯特理工学院计算机科学系,特洛伊,纽约,美国);Network Science and Technology Center, Rensselaer Polytechnic Institute, Troy, NY, USA(罗切斯特理工学院网络科学与技术中心,特洛伊,纽约,美国)