分类器选择的IDS系统入侵检测方法

摘要

为了提高IDS入侵样本检测的速度,提出了分类器选择的入侵检测方法.该方法充分考虑到线性分类器在数据分类操作中的速度优势,以线性可分度判定的的结果为基础,灵活选择(非)线性分类器进行数据分类,较之单一的SVM分类方法,虽然增加了额外的线性判别的系统开销,却因此获得分类性能的极大提升.在KDD Cup99数据集上进行实验,实验结果表明,该方法在保证分类精度、漏报率、误报率等性能参数不受影响的前提下,有效提高了入侵样本的训练和检测速度.%To speed up the detection process of intrusion samples, taking the speed advantage of linear classifier in data classification into account, based on linear discriminant, a method of classifier selection is put forward.Compared with single SVM classifier, the method enhanced classification performance, although the extra overhead is added.The results by experimented on KDD Cup99 sets show the method can greatly improve the speed of sample training and testing, and ensure the accuracy, the false alarm rate and omission rate is not affected.