基于SVDD的半监督入侵检测研究

Semi-supervised Intrusion Detection Research Based on SVDD

提出了一种基于SVDD的半监督入侵检测算法.该算法利用少量有标记正常网络数据建立两个SVDD分类器,通过相互学习来挖掘未标记数据中的隐含信息,扩大有标记正常网络数据的数量.再利用所有已标记正常网络数据用不同的单分类方法建立多个单类分类器,通过集成学习的方法得到最终的分类器.实验表明,该算法具有良好的识别性能.

In this paper a new semi-supervised intrusion detection algorithm based SVDD is proposed to solve the problem which has only some labeled normal network data and lots of unlabeled data. In this proposed algorithm two SVDD elassitiers are built firstly respectively for the known labeled normal data, then some information under the unlabeled data is mined by the two classifiers. And the known labeled target data is enlarged by the co-learning. At last the enlarged labeled target data is used to train three classifiers, and the ensemble learning is used to get the final classifier. The proposed algo-rithm has better recognition performance.