Resilience Analysis of Various Communication Network Models

Abstract

Graph theory provides a foundation for modeling real-life networks of which there are several modeling methods to consider, including multilayer, multiplex, and multigraph. Additionally, the operations research discipline provides means to modify a network to study concepts such as flow and minimum-cost routing. With various model choices and structural modifications, perhaps there are unintended consequences. To explore this idea, we model three simulated organizations’ communication networks and how they withstand adversarial attacks. For each model, we delete vertices in accordance with their ranked betweenness centrality measures. At each deletion, we perform various measurements of interest associated with cost, latency, and fragmentation. We then repeat the attack with the same betweenness centrality values and measures of interest, but this time only deleting the ten vertices with the largest betweenness centrality measures. Upon completion of our attack schemes, we observe models which outperform the others when we consider either solely latency or fragmentation. Model preference also reveals itself when we consider all measures of interest in a single, holistic assessment. With only one significant difference in our model preference between our full-attacks and top-ten attacks of all networks, we conclude the choice of network model is important.