Vulnerability of interdependent infrastructures under random attacks

Most of today's critical infrastructure are in the form of interdependent networks with new vulnerabilities attributed to their interdependencies. Security and reliability attacks, which can trigger failures within and across these networks, will have dierent forms and impacts on interdependent networks. In this paper, we focus on random attacks in a two-layer interdependent network and quantify its vulnerability under two dierent types of such attacks: 1) single layer attack, 2) concurrent two-layer attack. We compare the vulnerability of the network given the two attack scenarios, to answer the question of whether one single attack in one layer of an interdependent network can be as severe as concurrent multi-layer attack? We define two quantitative metrics to evaluate the vulnerability of the interdependent network under a given attack, namely, long-term eect and critical times. The long-term eect represents the total number of aected nodes during a finite time window, while critical times capture the evolution of failure propagation over time. The impact of dierent types of network topologies, including Erdos-Renyi, Scale-Free, and Small-World, and dierent coupling scenarios between the layers, namely high and low intensity and random and designed coupling on the vulnerability of the network is studied. Our results show that two-layer attacks are more severe in most cases while a single attack in one layer can be more severe for certain scenarios of coupling and network topologies. This suggests that in interdependent networks severe attacks can be triggered using access to only one layer of the network if the network structure is vulnerable.