一种基于神经网络与证据理论融合的P2P业务感知模型

P2P Traffic Identification Model Based on Combination of Neural Networks with Evidence Theory

提出了一种基于神经网络与证据理论融合的P2P业务感知模型,该模型利用神经网络的非线性逼近能力和自学习能力,获取证据理论所需的基本概率值;并通过证据理论的数据融合明显提高业务感知准确率。实验结果表明,该模型与现行的P2P业务识别方法相比,能够快速、准确、可靠地识别P2P业务类别,实现合法有效的网络管理和控制,对检测网络异常行为与提高网络安全性具有重要意义。

In order to improve the currently Internet traffic identification,we propose a novel P2P traffic identification model based on combination of neural networks with evidence theory.The model using neural networks nonlinear approximation ability and self-learning ability can be more objectively access to basic probability value required for evidence theory identification phrase,and improve greatly the Internet traffic identification accuracy after the re-integration of evidence theory.Contrasted with currently classification approaches,this model can identify Internet traffic with high efficiency and accuracy.It is very important to legal network management and control,anomaly detection and control and network security.