Comparative evaluation of strategies for improving the robustness of complex networks

Designing network systems able to sustain functionality after random failures or targeted attacks is a crucial aspect of networks. This paper investigates several strategies of link selection aiming at enhancing the robustness of a network by optimizing the effective graph resistance. In particular, we study the problem of optimizing this measure through two different strategies: the addition of a non-existing link to the network and the protection of an existing link whose removal would result in a severe network compromise. For each strategy, we exploit a genetic algorithm as optimization technique, and a computationally efficient technique based on the Moore–Penrose pseudoinverse matrix of the Laplacian of a graph for approximating the effective graph resistance. We compare these strategies to other state-of-the art methods over both real-world and synthetic networks finding that our proposals provide a higher speedup, especially on large networks, and results closer to those provided by the exhaustive search.