【摘要】 传统协议识别算法无法适应当前多变复杂的网络环境,尤其在当今复杂网络环境中P2P应用中广泛的动态端口应用。因此针对传统端口识别方法的局限性,提出了一种基于NetFlow的应用协议半监督识别算法。采用NetFlow报文数据为基础,通过对NetFlow属性维进行子空间聚类,生成每种协议类型的关键维组,采用半监督算法,根据每种协议的关键维特征识别NetFlow数据。实验结果表明,文中提出的基于NetFlow的应用协议半监督识别算法在测量准确度上优于传统识别算法,有效地解决了当前复杂网络中的协议识别问题。更多还原

【Abstract】 The traditional protocol recognition algorithms cannot adapt to the current complex and changing network environment,especially in the P2P applications,which use a wide range of dynamic port.Therefore,the limitation of the traditional port classification method,presents the semi-supervised recognition algorithm based on NetFlow.With NetFlow technology,through the sub-space clustering,select the main properties set,using semi-supervised algorithm on NetFlow data for identification application protocol.The results show that semi-supervised clustering algorithm for application protocol recognition base on NetFlow,the measurement accuracy is better than the traditional recognition algorithms to effectively solve the problem of today’s complex network protocols recognition.更多还原