Anomaly Detection Based on Network Module Structure
基于网络模块化结构的异常发现

The large scale and high speed networks create massive data and have low detection accuracy. To address the problems, the idea “module” is brought from complex network into anomaly detection area. Firstly, the relations between network partition strategy and network detection accuracy are modeled, and a theoretically proof is given that partition strategy which based on network modularity is favorable for anomaly detection. Secondly, the module-based detection is proved that has higher detection rate and efficiency than network-based detection by theoretical analysis and experiments. Finally, by using flow-splitting and parallel processing technologies this approach can improve efficiency obviously.