Optimizing Algorithm For Networks Robustness Under Limited Knowledge Attacks

Many realistic networks are found to have community structure, which plays an important role in realworld applications, therefore it is critical to maintain the functional integrity of communities when networks are subjected to attacks and errors. Plenty of previous studies have focused on the robustness of community structures against malicious attacks, in which the full information of all nodes in the network is wellknown. However, restricted by realistic conditions, the information obtained by observing networks is limited, so we adopt a novel attack strategy that only partial nodes are observed, and the most vulnerable nodes among these are targeted in each attack round, i.e., the limited knowledge attack. Based on this attack strategy, we designed an optimization algorithm MA-CRLA which improves the community robustness under limited knowledge attacks while keeping the initial community partition well-preserved. The results show that the algorithm can effectively enhance the community robustness under different attack strategies, and prove that even if only few numbers of nodes are observed in each round of attack, the limited knowledge attack can still achieve the destructive effect of an approximately malicious attack.