几种机器学习方法在IDS中的性能比较

Performance Comparison of Several Machine Learning Methods for Intrusion Detection

入侵检测是一种保障网络安全的新技术,传统的入侵检测方法存在误报漏报及实时性差等缺点,将机器学习的技术引入到入侵监测系统之中以有效地提高系统性能具有十分重要的现实意义。将目前主要的基于机器学习的贝叶斯分类的方法、神经网络的方法、决策树方法与支持向量机的方法应用于入侵检测系统中,以kdd99公共数据集进行了仿真实验,仿真测试结果表明支持向量机方法(SVM)和神经网络方法具有较好的分类识别性能,适合用于入侵检测。

Intrusion Detection Method is a new emerging network security technology.The traditional intrusion detection systems have high false negative rate,so it is important to introduce machine learning into intrusion detection systems to improve the performance.In this paper,currently popular machine learning methods including the Bayes method,the neural network method,the decision tree method and the Support Vector Machines（SVM） method are applied to intrusion detection system,experiments with the data set kdd99 show that the method SVM and the neural network method have better performance and are more suitable for intrusion detection.