关键词: 网络流量, 异常检测, 流时间影响域, 流量网络模型, 网络平均度

Abstract: Aiming at improving the accuracy rate of anomaly network traffic detection， a network traffic detection model was proposed based on the time influence domain(TID)of network flow. By analyzing the changes of average degree of traffic network model under the normal and abnormal conditions， an anomaly detection algorithm of network traffic based on the average degree metric of complex network was developed to detect the abnormal traffic. Experimental results show that based on the flow time influence domain， the anomaly detection model of traffic network can reasonably describe the inter-dependency relationship between network traffic. The proposed method has a better detection performance， meanwhile only three network features， i.e. timestamp， source IP and destination IP， are needed to implement the above model. Detection efficiency is better than other methods. The method proposed meets most network types and has a better ubiquity.

Key words: network traffic, anomaly detection, flow time influence domain, traffic network model, network average degree