基于CNN与WRGRU的网络入侵检测模型  

Network Intrusion Detection Model Based on CNN and WRGRU

作　　者：王运兵[1] 姬少培 查成超 WANG Yunbing;JI Shaopei;ZHA Chengchao(No.30 Institute of CETC,Chengdu Sichuan 610041,China)

机构地区：[1]中国电子科技集团公司第三十研究所,四川成都610041

出　　处：《通信技术》2022年第4期486-492,共7页Communications Technology

基　　金：国家自然科学基金企业创新发展联合基金(U19B2021)。

摘　　要：针对当前的入侵检测方法普遍存在准确率与泛化性较低的问题,提出了一种基于卷积神经网络(Convolutional Neural Network,CNN)和权重缩减门控循环单元(Weight Reduction Gated Recurrent Unit,WRGRU)的网络入侵检测模型(CNN-WRGRU)。该模型首先利用CNN进行入侵检测数据集的特征提取,其次利用WRGRU来学习数据特征之间的依赖关系,保留了特征之间的长期相关性,有效地防止了过拟合现象的出现,提高了模型的识别准确率及泛化性。在实验中,将CNN-WRGRU与传统方法在公开数据集上进行了检测性能比较,结果证明CNNWRGRU模型具有更好的识别效果,有效地提高了入侵检测的识别精度。Aiming at the problems of low accuracy and generalization in current intrusion detection methods,this paper proposes a network intrusion detection model(CNN-WRGRU) based on CNN(Convolutional Neural Network) and WRGRU(Weight Reduction Gated Recurrent Unit). The model first uses CNN to extract features from the intrusion detection dataset, and then uses WRGRU to learn the dependencies between data features, retains the long-term correlation between features, effectively prevents the occurrence of over-fitting, and improves the recognition accuracy and generalization of the model. In experiments, the detection performance of CNN-WRGRU is compared with traditional methods on public datasets. The results indicate that the method proposed in this paper has better recognition effect and effectively improves the recognition accuracy of intrusion detection.

关 键 词：入侵检测 权重缩减门控循环单元 特征提取 依赖关系 识别精度 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]