CORRELATION ANALYSIS OF ATTACK STRATEGIES AND ROBUSTNESS ANALYSIS ON THREE NETWORK MODELS

Network centrality metrics based on heuristics which work as a guideline on removing vertices is a vital topic for the research of network robustness. In this paper, we analyze the correlation and difference between 12 popular centrality metrics among three distinct network models, including the correlation analysis and granularity anal-ysis. We also concern about the destructiveness of these centrality metrics when they are used as attack strategies. The results show that most of the centrality metrics are highly correlated with each other across three network models. The granularity analysis on centrality metrics is also considered in this paper. The experiment results also ver-ify the previous conclusion that scale free networks are more vulnerable to intentional attack than other networks. We observe that more correlated centrality metrics would cause more similar damage to network connectivity; at the meantime, we also find that the attack strategies based on few centrality metrics could destroy networks very quickly. All these results inspire us that there is a paradigm that could detect the attack strategy with strong destructiveness for distinct networks by combining the highly correlated and highly destructive metrics.