Comparing Destructive Strategies for Attacking Networks

Abstract

The failures of multiple elements in a network can have disastrous consequences on its operation. Therefore, understanding the robustness of networks that experience multiple failures is utterly important. In this chapter, we review well-defined metrics related to the topology and resilience of the network and use them to analyze the robustness of real-world networks under multiple failures. We consider 52 real-world networks from three different infrastructure domains, namely metro networks, power grids and telecommunication networks. We quantify the impact of targeted node removals from a network on the relative size of the largest connected component of the network. Nodes are attacked according to traditional centrality metrics, such as degree, betweenness, closeness and the principal adjacency matrix eigenvector. In addition, we consider attacks based upon the recently proposed ``zeta-vector'', that is the diagonal elements of the pseudo-inverse of the Laplacian matrix. Finally, we compare and rank these node-removal strategies and apply to the selected set of real-world infrastructures.