Network Under Joint Node and Link Attacks: Vulnerability Assessment Methods and Analysis

Critical infrastructures such as communication networks, electrical grids, and transportation systems are highly vulnerable to natural disasters and malicious attacks. Even failures of few nodes or links may have a profound impact on large parts of the system. Traditionally, network vulnerability assessment methods separate the studies of node vulnerability and link vulnerability, and thus ignore joint node and link attack schemes that may cause grave damage to the network. To this end, we introduce a new assessment method, called β-disruptor, that unifies both link and node vulnerability assessment. The new assessment method is formulated as an optimization problem in which we aim to identify a minimum-cost set of mixed links and nodes whose removal would severely disrupt the network connectivity. We prove the NP-completeness of the problem and propose an O(√{logn}) bicriteria approximation algorithm for the β-disruptor problem. This new theoretical guarantee improves the best approximation results for both link and node vulnerability assessment in literature. We further enhance the proposed algorithm by embedding it into a special combination of simulated annealing and variable neighborhood search method. The results of our extensive simulation-based experiments on synthetic and real networks show the feasibility and efficiency of our proposed vulnerability assessment methods.