Community Cut-off Attack on Malicious Networks

This paper aims to provide an efficient algorithm for quick disabling of known malicious network by sequential removal or incapacitation of their nodes. The nodes are selected for deletion in such a sequence, that the network is swiftly separated into small disjoined parts. We propose using a community detection based on random walks. For all the divisions of the found communities into two separate sets we create bigraphs defined by the edge set with each edge's node in different community and use Koenig's theorem to find the best vertex cut (set of vertices to be deleted). This community detection and their separation is used recursively on a currently maximal component of the network. The effectiveness of our algorithm is tested on both real-world and model networks by quantifying network robustness measure R based on the size of maximum component. Its results compare favorably against standard centrality based attack strategies.