利用统计特征结合神经网络的P2P流量识别方法

准确识别P2P流量对进一步地流量控制具有重要的实际意义。利用模糊ARTMAP神经网络实时学习和快速识别的优点,提出一种基于神经网络的P2P流量识别方法。在实际网络环境下对BitTorrent、PPLive、PPStream、EMule四种主流P2P应用进行实验,统计分析并提取了九种流量特征。通过神经网络对各种P2P应用流量特征的学习和识别,得出该方法的识别准确率达到95%以上,验证了其有效性。

Identification of P2P Traffic Using Statistics Method and Neural Network

Identify P2P traffic accurately is necessary to further flow control and it has an important practical significance.In this paper,we proposed one P2P traffic identification method based on fuzzy APTMAP neural network taking advantage of real-time learning and rapid recognition.In real network environment,we did statistical analysis of four mainstream P2P application which are BitTorrent、PPLive、PPStream and EMule,and extract their nine flow characteristics.After the neural network did its real-time learning and rapid recognition to every kind of P2P application＇s flow characteristics,we get a high accuracy identification rate which reaches 95% above that verifies validity of the method we mentioned above.