Optimizing Networks against Malicious Attacks
We study the robustness of different types of networks under dynamic degree-based node attack (DDA). The DDA removes the node with the highest connectivity and recalculate the connectivity at each removal step.
We optimize numerically the network against this attack strategy, under the condition that the degree distribution remains constant. During the optimization process we calculate some properties of the network, like clustering coefficient, assortativity, correlation, efficiency, robustness against random errors and robustness against other attack strategies to see what properties are important for robust networks. We find that the robustness against a specific attack strategy can be improved dramatically.
Fraction of nodes in the largest cluster as a function of the fraction of the nodes removed from the network under DDA. The results are the average over 100 random networks. We see data for 128 nodes before optimization (dotted line) and after optimization (straight line) for a scale free network (P (k) ∼ k −γ ). It is evident that the optimization can greatly increase the robustness of the network.