基于朴素贝叶斯算法的VoIP流量识别技术研究

摘要

近年来互联网技术飞速发展，实时通讯类应用的种类不断增加，这使得合理有效的管理网络流量面临着巨大的挑战。文章首先对VoIP这类网络流量使用的相关协议做了概述，然后分析了传统的流量识别方法，在此基础上，对于UDP流量或者加密流量提出基于传输层的识别机制，并且将朴素贝叶斯分类算法应用于对VoIP流量的识别，通过给出一个部署方案，对几个VoIP应用进行了验证。实验结果表明，朴素贝叶斯算法应用于VoIP流量分类可以取得很好的识别效果。最后针对朴素贝叶斯算法的属性独立性的约束，提出了该算法仍需要改进的地方。%With the development of Internet technology and the surge for the number of real-time communications applications, to effectively manage network traffic has been faced with new challenges.The article ifrst outlines protocols associated with VoIP network trafifc.Then on the basis of analyzing the methods for traditional traffic identification, it proposes an recognition mechanism based on transport layer for UDP trafifc or encrypted trafifc.And the article applies the naive Bayesian classifier algorithm to identification of VoIP traffic. Finally, it gives a deployment to verify several VoIP applications. Experimental results show that naive Bayesian algorithm is applied effectively to classiifcation for VoIP trafifc. At the end of the article,it describes that the algorithm still needs to be improvemented because of properties independence.