关键词: 流量分类, 聚类算法, 高斯混合模型, 最小描述长度（MDL）准则

Abstract: Network traffic classification is elementary to network security and management. With the development of network and information technology, the limitation of traditional port-based and payload-based classification approaches is that they can not classify network traffics accurately. This paper proposes a new semi-supervised approach based on correlation of flows, which is formulated by an MDL-CON Gaussian mixture model. In the process of cluster, the correlation between different flows is used to improve the quality of resultant traffic clusters, and the MDL rule is applied to solve preset clusters number and initialization issue. Experiments show that this approach can significantly improve the accuracy of traffic classification.

Key words: traffic classification, cluster algorithm, Gaussian mixture model, Minimum Description Length（MDL） criterion