• 首页
  • 期刊简介
  • 编委会
  • 投稿指南
  • 收录情况
  • 杂志订阅
  • 联系我们
引用本文:葛继科,刘浩因,李青霞,陈祖琴.基于改进CNN-LSTM的网络入侵检测模型研究[J].软件工程,2022,25(1):56-58.【点击复制】
【打印本页】   【下载PDF全文】   【查看/发表评论】  【下载PDF阅读器】  
←前一篇|后一篇→ 过刊浏览
分享到: 微信 更多
基于改进CNN-LSTM的网络入侵检测模型研究
葛继科,刘浩因,李青霞,陈祖琴
(重庆科技学院智能技术与工程学院,重庆 401331)
gjkweb@126.com; lhaoyin@qq.com; 690887086@qq.com; chenzuq81@163.com
摘 要: 针对网络入侵检测模型特征提取算法复杂、训练参数过多、检测结果不理想等问题,提出一种改进卷积神经网络与长短期记忆网络结合的网络入侵检测方法(GCNN-LSTM)。首先,使用卷积神经网络对流量数据做特征选择,并选择全局池化层代替其中的全连接层;其次,结合长短期记忆网络强大的时间序列学习能力对改进卷积神经网络选择后的特征进行学习分类,以期在网络异常数据检测方面获得更好的效率和准确率。实验结果表明,提出的模型在UNSW-NB15数据集上有着较好的检测效果。在同等条件下,使用传统卷积神经网络的模型准确率为84.97%,训练时间为76.3 s;本模型准确率达到了88.96%,训练时间为61.1 s。
关键词: 卷积神经网络;LSTM;全局池化;网络入侵检测
中图分类号: TP393.8    文献标识码: A
基金项目: 重庆科技学院研究生科技创新项目(YKJCX2020816);重庆市高等教育教学改革研究项目(202078).
Research on Network Intrusion Detection Model based on Improved CNN-LSTM
GE Jike, LIU Haoyin, LI Qingxia, CHEN Zuqin
(School of Intelligent Technology and Engineering, Chongqing University of Science and Technology, Chongqing 401331, China )
gjkweb@126.com; lhaoyin@qq.com; 690887086@qq.com; chenzuq81@163.com
Abstract: Aiming at the problems of complex feature extraction algorithm, too many training parameters, and unsatisfactory detection results in the network intrusion detection model, this paper proposes a network intrusion detection method (GCNN-LSTM) combining improved convolutional neural network and long short-term memory (LSTM) network. Firstly, convolutional neural network is used to perform feature selection on the flow data, and its full connection layer is replaced by global pooling layer. Then, in view of its powerful time series learning ability, LSTM is used to learn and classify the features selected by the improved convolutional neural network, in order to obtain better efficiency and accuracy in network abnormal data detection. Experimental results show that the proposed model has a good detection effect on the UNSW-NB15 dataset. Under the same conditions, the accuracy of the model using the traditional convolutional neural network is 84.97%, and its raining time is 76.3 s, while the accuracy of the proposed model is 88.96%, and its training time is 61.1 s.
Keywords: convolutional neural network; LSTM; global pooling; network intrusion detection


版权所有:软件工程杂志社
地址:辽宁省沈阳市浑南区新秀街2号 邮政编码:110179
电话:0411-84767887 传真:0411-84835089 Email:semagazine@neusoft.edu.cn
备案号:辽ICP备17007376号-1
技术支持:北京勤云科技发展有限公司

用微信扫一扫

用微信扫一扫