MassExodus: modeling evolving networks in harsh environments

Consider networks in harsh environments, where nodes may be lost due to failure, attack, or infection—how is the topology affected by such events? Can we mimic and measure the effect? We propose a new generative model of network evolution in dynamic and harsh environments. Our model can reproduce the range of topologies observed across known robust and fragile biological networks, as well as several additional transport, communication, and social networks. We also develop a new optimization measure to evaluate robustness based on preserving high connectivity following random or adversarial bursty node loss. Using this measure, we evaluate the robustness of several real-world networks and propose a new distributed algorithm to construct secure networks operating within malicious environments.