Traffic Identification in Semi-known Network Environment

Network traffic classification has attracted more and more attentions from both academia and industry. It has been widely adopted in network management and security, such as QoS measurements. Due to rapid emergence of new applications in current network environment, it is impractical for a classification system to obtain full knowledge of a network environment. A big challenge to the identification of interested traffic comes from semi-known network environment, in which some emerging applications are not recognized by the classification system yet. In this paper, we proposed a new framework of Traffic Identification with Unknown Discovery (TIUD) by innovatively combining supervised and unsupervised machine learning techniques to meet the challenge. The proposed TIUD framework has the capability to accurately identify the interested traffic in semi-known network environment. The proposed framework is fully evaluated on a large real-world traffic dataset, with a comparison with three state-of-the-art traffic classification methods. The experimental results yield a outstanding performance of the proposed framework.