一种基于扩充特征集的流分类方法

Traffic classification based on extended feature set

鉴于当前流分类研究均建立在使用载荷无关的流特征的基础上,而载荷无关的特征一般无法为准确分类提供充足的分类信息的问题,提出了一种基于扩充特征集的流分类方法,该方法首先提取载荷特征扩充流分类特征集合,特征集合扩充后,特征的数目显著增加,呈现出高维特性,进而针对高维特征空间,提出了一种混合型特征选择算法,并基于该算法选取的特征构建流分类器。实验结果表明,相对于使用载荷无关特征集的方法,所提出的方法能够显著改善分类效果,同时能够提升分类速度,更适用于现实网络环境。

In consideration of the problem that the present studies on traffic classification are all based on the use of payload- independent features, but the payload-independent features often do not contain sufficient information to allow for an accurate methodology, the paper proposes a traffic classification method based on the extended feature set. The method extends the flow feature set with payload, and after extension, the feature number is significantly increased. For high dimension data, the paper then proposes a hybrid feature selection algorithm for traffic classification and builds the classifier with the selected features. The experimental results demonstrate that the propesed method can not only guarantee a high classification accuracy but also a better performance in terms of classification speed. Therefore, it is more suitable for the real network applications than the traditional approaches.