| Summary: | 碩士 === 逢甲大學 === 資訊工程學系 === 101 === In recent years, due to the rapid development of the P2P applications, the P2P network traffic occupies more than 50% of the overall network traffic. The users will occupy most of network bandwidth when using P2P file-sharing software which will affect other user’s network speed and reduce the overall network performance. Therefore, for network engineers, it is necessary to identify and control the P2P traffic effectively.
Many techniques used to avoid detection have been proposed such as using dynamic port number, encrypt payload, etc.; these techniques make the traditional traffic classification methods gradually unable to meet the demand of today’s traffic classification. Due to the diversity of the P2P applications, high false classification rate may occur if we only use single classification method to classify the P2P flows. Hence, this thesis proposes a multi-level P2P classification system. The system integrates multiple classification techniques to combine their advantages. It includes the fast preliminary classification and the long-term classification achieving to classify most flows fast with a low false classification rate. In addition, it also has the feedback mechanism that can let more flows be classified fast by feedback results to improve the classification speed.
|
|---|
