基于信息增益的贝叶斯入侵检测模型优化的研究

The Research of an Optimized Bayesian Intrusion Detection Model Based on Information Gain

对于不同类型的网络入侵,其行为特征所涉及到的主要数据属性会有所不同。传统的朴素贝叶斯(NB)入侵检测模型没有考虑这个差别。本文引入信息增益改进传统的NB模型,利用它来对网络连接数据的属性进行特征选择,并删除一些冗余的属性,达到优化NB入侵检测模型的目的。实验表明,信息增益对NB模型有一定的优化作用,相对神经网络模型有更高的检测率。

As for different network intrusion detections, their different actions have different data attributes. The traditional Naive Bayesian（NB） model of intrusion detection did not consider the difference. The paper exploits information gain in order to improve the traditional NB model,and use it to select features and delete unneeessary attributes in order to optimize NIK The experimental results show that information gain can optimize the traditional NB model to some extent, and have a higher detection rate for neural networks.