Internet application identification is needed by network management in many aspects,such as quality of service (QoS) management,intrusion detection,traffic engineering,accounting,and so on. This article makes an in-depth study of precise identification of Internet applications by using flow characteristics instead of well-known port or application signature match. A novel approach that identifies the application type of an Internet protocol (IP) flow by finding what flow the flow looks the most like based on medium mathematics system (MMS) is proposed. The approach differs from previous ones mainly in two aspects:it has inherent scalability due to its use of the measure of n-dimensional medium truth degree; not only features of a flow,but also the association between the flow and the other flows of the same host as well as the relation among all flows of a host are employed to recognize a flow's application type. For the present,some popular applications are concentrated on,and up to six application types can be identified with better accuracy. The results of experiments conducted on Internet show that the proposed methodology is effective and deserves attention.
