Research On Network Traffic Classification Based On FPGA

With new technologies such as P2P, VOIP emerging and complexity of network application is growing, the nature of network traffic has been innovative. Through effective technical means to manage and control network traffic, to provide different levels of service quality is challenges that the current network operators facing.The background and development of research for network traffic classification has been introduced in this paper. The method of identifying network traffic through port number is very simple, but can not identify applications with dynamic ports. Relying on the inspection of packet contents, payload-based techniques have been proposed. Transport layer statistic techniques have been motivated by the limitations and disadvantages of port-based and payload-based approaches such as identifying new P2P application.In this paper, a new approach based on the implementation of artificial neural network ensemble with ECOC (Error-correcting output codes) is proposed for classification of multi-class network traffic. ECOC is a method of dividing multi-classification problem into many binary classification problems. It improves the generalization performance of the base classifiers. Moreover, the ECOC technique decreases the error caused by the bias and the variance of the base learning algorithm. As the error correcting output codes have error correcting ability and improve the generalization ability of the base classifiers, the experiments show that the proposed method can improve the multi-class classification accuracy dramatically.However most of the proposed algorithms are off-line and one classifier model, designed for online real-time network traffic classifier is clearly necessary. An effective and intelligent on-line identification method is becoming focus in this field. In this paper, a new hierarchical real-time model is proposed in the first time, in which there is a combination of a three tuple (source IP, destination IP and destination port) look up table and layered milestone ECOC (Error Correcting Output Codes) model. This algorithm is specially designed for the traffic classification problem, and does not apply to other classification problems. Experiments show that the proposed model can improve the real-time efficiency better than other methods. FPGA-based network traffic classification system consists two parts: FPGA side and PC side. Data packet reception, extracting the five-tuple of a packet and determining whether the data packet has created a new data stream, extracting and updating the characteristics of each data flow and sending them to PC will be implemented in FPGA. PC terminal receives characteristics from FPGA and send them to hierarchical real-time model, then monitor the composition and distribution of network.The study of network traffic classification has been completed systematically.