代价敏感支持向量机在入侵检测中的应用

The application of Intrusion detection using Cost-sensitive SVM

入侵检测系统在最大化计算机安全性的同时,着手减小其代价也是关键点之一。标准的分类器设计一般基于精度,在入侵检测等实际应用问题中,不同的类别对应的错分代价也不同,在此类问题中直接使用标准分类方法就无法取得良好的分类和预测效果。代价敏感算法通过改变代价矩阵,可使高代价样本的错分率得到有效的控制,并尽量减少总体错分代价。本文对代价敏感支持向量机在入侵检测中的应用进行了研究,并用KDDCUP99标准数据集对文中算法进行了测试评估。

While maximizing the safety of computer by Intrusion detecting system （IDS）,minimize the cost is also one of the key point. Standard classifiers normally base on minimizing the incorrectly classified error; however,in some applications like intrusion detection,different misclassification has different cost. Therefore,using traditional classification methods on these areas will not get the best classifying and predicting effect. Cost sensitive algorithm could adjust misclassification rate of high cost sample by changing the cost matrix,and try to minimize the general misclassification cost. This paper has a research on the application of Intrusion de-tecting using cs-svm,and KDDCUP99 dataset is used to test.