Topological Fragility Versus Antifragility: Understanding the Impact of Real-Time Repairs in Networks Under Targeted Attacks

摘要

Network science fosters our understanding of many natural and man-made complex systems by providing models for evolution, emergence and reliability. Existing studies on network robustness mainly focus on targeted node attacks, where the topology shrinks until the network loses its basic functionality. Here, we bring a novel aspect into perspective. First, we investigate the fragility of networks when dealing with edge attacks instead of node attacks - a highly relevant scenario in dynamical social systems where, because of the frequent interactions, edges are dynamically created or destroyed. Second, we propose an intuitive edge repair model to respond after edge attacks. Accordingly, we study different strategies for targeted edge attack and edge repair, to find that real-world networks are generally more robust than synthetic ones, that high betweenness and high degree nodes present structural weak-spots for targeted attacks, and that repairing edges incident to high-degree nodes first offers the most effective response to an attack. Finally, we find evidence of a behaviour we tag as topological antifragility for networks that actually grow stronger when being under attack, due to the triggered edge repair strategy.