中文摘要：

针对现有入侵检测算法中存在着冗余或噪声特征导致的检测模型精度下降与训练时间过长的问题进行了研究,将特征选择算法引入到入侵检测领域,提出了一种基于特征选择的入侵检测方法。利用不同的离散化与特征选择算法生成具有差异的多个最优特征子集,并对每个特征子集进行归一化处理,用分类算法对提取后的特征进行学习建模。通过实验将该方法与基于传统算法（决策树、朴素贝叶斯、支持向量机）的入侵检测方法作比较,实验结果表明,该方法有效地提高了检测攻击的准确率,并且降低了模型的训练时间。

英文摘要：

The intrusion detection system deals with huge amount of data which contains redundant and noisy features causing poor detection rate and slow training process. This paper introduced feature selection algorithm into the field of intrusion detec- tion, and put forward a intrusion detection method based on feature selection. It used different discretization and feature selec- tion algorithm to extract difference of multiple optimal feature subset, followed by normalizing the extracted feature subsets to perform a normalizing process. At last it applied the classification algorithm to create a model. Compared with the traditional algorithm （decision tree, naive Bayes, support vector machine） , the experimental results demonstrate that the approach can effectively improve the precision of attack-detection and training cycle.