结合堆栈自编码器和FSVM的入侵检测方法  

Intrusion detection method combining stacked autoencoder and FSVM

作　　者：张海洋 张斌 ZHANG Haiyang;ZHANG Bin(School of Computer and Information Security,Guilin University of Electronic Technology,Guilin,541004,China;School of Electronic Engineering and Automation,Guilin University of Electronic Technology,Guilin,541004,China)

机构地区：[1]桂林电子科技大学计算机与信息安全学院,广西桂林541004 [2]桂林电子科技大学电子工程与自动化学院,广西桂林541004

出　　处：《桂林电子科技大学学报》2024年第6期621-627,共7页Journal of Guilin University of Electronic Technology

基　　金：国家自然科学基金(61762028)。

摘　　要：大规模、高维度的样本数据会对支持向量机模型的学习速度与学习效果产生巨大影响,为此提出一种结合堆栈自编码器和快速支持向量机的入侵检测方法。搭建基于堆栈自编码器的降维模型,通过该模型对网络数据进行特征提取,获取高质量的特征数据。使用聚类算法分别对样本中不同类型的数据进行聚类操作,计算不同类型样本间簇的相似度,以获取不同类型样本间的相似簇;将相似簇中的样本数据作为支持向量机模型的输入,从而降低输入样本的规模,提高模型的学习速度。仿真结果表明,本方法可有效提高支持向量机模型对大规模高维数据的学习速度和学习效果。Large-scale,high-dimensional sample data have a strong impact on the learning speed and learning effect of the l,so an intrusion detection method combining stacked auto-encoder and fast support vector machine(SAE-FSVM)was come up for discussion.A dimensionality reduction model based on stacked auto-encoder was built,through which the network data was extracted with features to obtain high-quality feature data.The clustering algorithm was used to cluster different types of data in the samples separately.The similarity of clusters between different types of samples was calculated to obtain similar clusters between different types of samples,and the sample data in the similar clusters were used as the input of the support vector machine model,so as to reduce the scale of the input samples and improved the learning speed of the model.Through simulation experiments,it is verified that this method can effectively improve the learning speed and learning effect of the support vector machine model for large-scale high-dimensional data.

关 键 词：支持向量机(SVM) 特征提取 样本规模归约 堆叠自编码器 入侵检测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]