Dynamic online traffic identification scheme based on data stream clustering algorithm

Although researches on network traffic identification have already got some achievements, but most of them are not suitable for online traffic classification by considered the dynamic feature of flows. In this paper, we propose a dynamic online traffic identification method by introducing density-based clustering algorithm for stream data called DStream, and using the feature select algorithm to reduce dimension of feature set. The result of classification is compared with CluStream algorithm. The classifier shows better performance than CluStream, and outliers can be detected.