Procedures for breaking networks are attracting much attention from two points of view: attacking and preventing attacks or failures. In this contribution we present a novel procedure to break complex networks [3,5] guided by the identification of modular structures. Our method first identifies communities [2,4] in which the network can be represented, then it deletes the nodes or edges that connect different modules by decreasing order in the betweenness centrality ranking list. We illustrate the method by applying it to various examples of real networks from social, infrastructure, and biological context.

We compare the fragmentation procedure with several previous and well accepted criteria, showing that our procedure always performs better than attacks based solely on the ranking of centrality measures. Improvement in efficiency of the attacks are remarkable with gains that increase exponentially as the modularity of the networks increases. As an example, for the US power grid [1] the present method breaks the original network of almost 5000 nodes in many fragments less than 4% of the original size by removing less than 3% of the total number of nodes. In contrast, with the same amount of nodes removed by degree or centrality based procedures, the power grid remains with 82% of its original structure.


Sebastian Goncalves Associate Professor

Open e-session

Photos by :