Robustness of Real Network Controllability to Degree Based Attacks




Journal Title

Journal ISSN

Volume Title



Real world complex networks vary greatly topologically from each other as well as from generated synthetic random networks. For example, social and biological networks typically have a community structure, while random networks do not. In order to investigate the robustness of the controllability of real networks to attacks on its edges, five different attacks based on the degree of the nodes were levied at common real-world networks systematically by removing either 2% or 5% of the edges, in steps, until 90% were removed. It was then investigated how well these real networks retain their controllability, especially in comparison to Erdos-Renyi and Barabasi-Albert synthetic networks. In particular, the question of how effective attacks focusing on destroying edges with a high source node in-degree and a high target node out-degree, performed in comparison to attacks focused on edges with high betweenness centrality, was reviewed. It was discovered, that in contrast to results with synthetic networks, for many real networks the betweenness attack performed worse than the in-out attack after a certain number of edges were removed. By observing how high density and community structure affect the ability to retain control over the network after these two attacks, an explanation for this may be assembled. In addition, the difference between the potency of these two attacks, while network controls were fixed to nodes or allowed to move to more optimal input nodes, was studied.



Graph theory, Robust control, Control theory, Betweenness relations (Mathematics)