摘要
为了提高网络入侵检测(NID)系统的检测准确度,适应现代网络需求,提出一种入侵检测的深度学习方法。该方法利用堆叠式非对称深度自编码器(NDAE)构建深度学习分类模型,将堆叠式NDAE(深度学习)和随机森林(浅层学习)的优点相结合,以支持NID在现代网络中的运行。实验使用KDD Cup’99和NSL-KDD基准数据集对所提分类器进行评价。实验结果证明了所提方法的有效性,其分类器能够有效降低网络入侵检测的时间,精简数据特征,提高检测精度,实现了最高约5%的召回率提升和最高98.81%的训练时间缩减。
In order to improve the detection accuracy of network intrusion detection(NID) system and meet the needs of modern network, a deep learning method of intrusion detection is proposed. In this method, the stack non-symmetric deep auto-encoder(NDAE) is used to build a deep learning classification model, which combines the advantages of stack NDAE(deep learning) and random forest(shallow learning) to support the operation of NID in modern networks. In the experiment, KDD cup’99 and NSL-KDD benchmark data sets are used to evaluate the proposed classifier. The experimental results show the effectiveness of the proposed method. The classifier can reduce the time of NID, simplify the data features, and improve the detection accuracy. The classifier achieves a recall rate increase of up to 5% and a training time reduction of up to 98.81%.
作者
刘炜
LIU Wei(School of Information Engineering,Shaanxi Xueqian Normal University,Xi'an,710100,China)
出处
《控制工程》
CSCD
北大核心
2021年第9期1879-1885,共7页
Control Engineering of China
基金
陕西省提升公众科学素质计划(2021)项目(2021PSL39)
教育部产学合作同育人项目(201802146010)。
关键词
网络入侵检测
深度学习
非对称深度自编码器
分类器
随机森林
Network intrusion detection
deep learning
non-symmetric deep auto-encoder
classifier
random forest
