Recognition Model Of Network Traffic And Achieve

Since the Internet entered China in last mid-nineties, its business developed in full swing, although it experienced ups and downs, but after all, has brought to mankind a new communication tools and approaches. Internet-based applications grows rapidly like an explosion, caused a rapid expansion of network traffic. Application development is accompanied by the threaten , network quality and network security are growing concerned by users. Network traffic identification and classification techniques have become increasingly important, it is an important basis for Network management, network security, and quality of service。Research traffic identification and classification technology to improve the identification and classification speed, accuracy and the ability to accurately identify the Unknown traffic capacity is becoming particularly important.This paper provides an overview of the traffic recognition principles, key technologies and monitoring methodologies. In conjunction with the research background in this field ,present my own research methods, and validate the method mentioned in this paper. This research content and results are as follows:1. According to previous research, combining with internet application status, divide the Internet traffic into many classes.2. Fully and deeply analyze widely used network protocol ,analyze content includes protocol description documents, raw data packet captured by some capture software like Ethereal. Drawn common features of the each class of network traffic, or a certain type of an application protocol, or flow features. And to ensure that these characteristics can be accurately identify the type of traffic. To sum up these features for the flow identification .3. Improving the port-based identification method. By the conclusions of the analysis process, figure out the flows which used to be identified by port number, but currently changed. Through a number of additional conditions to confirm the class of traffic.4. For some of the latest applications which can not be classified by port-based method, analyze its payload which included the characteristics, by using information-based payload approach to identify these flows. In chapter III, we analyze some new applications which accounted for a large proportion of network traffic in recent years。For some flows those cannot be classified by port-based method ,it is necessary to analyze their payloads and draw features from their payloads.5. The Bayesian neural network learning function applied to the identification of network traffic. Pre-label the samples, and select a set of flow characteristics for learning, use learned Bayesian neural network to identify the traffic. It can identify not only the known traffic, but also the traffic generated by the unknown classification.6. By integrating port-based, payload-based, machine learning based methods ,to achieve a complementary effect, it indeed effectively improve the recognition rate and recognition accuracy.