Revisiting Models of Vulnerabilities of the Networks

Abstract

The study critically revisits the models of ‘vulnerability’ of various types of networks and shows several limits of the current state of the art, including the ambiguous definitions of vulnerability and robustness concepts, the arbitrary use of various connectivity indexes on graphs for assessing the ‘vulnerability’ of real-world networks to real-world attacks, the lack of any evidence for the proposed models, and the lack of significant statistical approaches. Next, the study shows the limits of using the world ‘measure’ and ‘metric’ in relation with connectivity characteristics on graphs. Then, a general cause-effect chain for attacks on networks, especially computer and transportation networks is laid down as a basis for probabilistic model building. Evidence is provided on the relation between graph features and the probabilities of events under an attack and model examples are discussed. An annex on the caution of making public detailed knowledge on such models ends the paper.