Traffic Evaluation Model For P2P Decision Algorithm Based On Distance

In recent years, new technologies emerging in the Internet, P2P, this new peer-to-peer network communication mode, a new application model as the Internet quickly swept the world. P2P technology has completely changed the mode of Internet applications, P2P traffic has occupied more than global Internet business. Such a huge traffic to communications carriers has caused tremendous pressure, how to control P2P traffic has become a huge challenge for operators are more efficient. Therefore, fast and efficient and accurate identification of P2P traffic is becoming an urgent problem.However, with the rapid development of peer-to-peer technology, P2P traffic have adopted new technologies such as dynamic port, protocol field encryption to evade detection of P2P traffic. Hide traffic technology, the traditional P2P traffic identification method has gradually become no longer applicable.In view of this, from the characteristics of P2P technology, the importance of traditional identification methods of P2P traffic and P2P traffic identification to start, according to the characteristic properties of P2P streaming the number of packets, UDP proportion and number of connections, combined with data mining and other classification algorithm, network data stream identification classification model based on distance judgment algorithm. Judgment parameters of the identification model consists mainly of number of UDP proportion (UDP protocol accounted for in the period of flow ratio of the flow of the paragraph all network protocols) and the number of connections (the same as the source IP address for a traffic destination IP address and of all connections) These two features properties in the processing of the number of connections characteristic attributes use of Java as a development language for the extraction of eigenvalues.11typical P2P network stream, and11non-P2P network flow characteristic attribute value extraction, established identify the model parameter values, and finally random crawl6groups of network flows in the LAN network flow through the identification model classification ruling the verdict known online behavior consistent verify the effectiveness of the network traffic classification recognition model.