半监督学习在网络入侵分类中的应用研究

Network intrusion classification based on semi-supervised learning

为了解决网络环境中已标记入侵数据获取代价大的问题,将半监督学习引入网络入侵分类领域。根据网络攻击类型的不同,将少量的已标记入侵数据分为三部分,分别作为最初的训练集训练分类器,形成三个差异较大的初始化分类器。通过三个分类器协同学习,实现对未标记入侵数据进行标记。详细介绍了使用KDD Cup99数据集构造半监督分类实验数据集的过程。实验结果表明,半监督学习能有效地挖掘未标记入侵数据信息,具有较高的入侵分类率。

In order to solve the problem that it costs too much to obtain labeled intrusion data in the network environment,semi-supervised learning is applied into the field of network intrusion.According to the different types of network attack,this paper divided the limited labeled intrusion data into three equal training sets to form three different classifiers.Through training learning by three single classifiers,the unlabeled samples were labeled.It introduced the process of using KDD Cup 99 data sets to construct semi-supervised classification experiment data sets.The experimental results show that semi-supervised learning can effectively dig the unlabeled samples information of intrusion data and has a higher rate of intrusion classification.