基于GR-CNN算法的网络入侵检测模型设计与实现  

DESIGN AND IMPLEMENTATION OF NETWORK INTRUSION DETECTION MODEL BASED ON GR-CNN ALGORITHM

作　　者：池亚平[1,2] 杨垠坦 李格菲 王志强 许萍[2] Chi Yaping;Yang Yintan;Li Gefei;Wang Zhiqiang;Xu Ping(School of Communication Engineering,Xidian University,Xi’an 710071,Shaanxi,China;Department of Communication Engineering,Beijing Electronic Science and Technology Institute,Beijing 100070,China)

机构地区：[1]西安电子科技大学通信工程学院,陕西西安710071 [2]北京电子科技学院通信工程系,北京100070

出　　处：《计算机应用与软件》2019年第12期297-302,333,共7页Computer Applications and Software

基　　金：国家重点研发计划项目(2018YFB1004101)

摘　　要：针对现有网络入侵检测系统对网络行为检测准确率较低、实时性较差、泛化性能较低的问题,利用深度学习具有良好分类性能及强泛化能力等优点,设计基于增益率算法和卷积神经网络算法的网络入侵检测模型。采用增益率筛选数据集数据特征,在保证入侵检测准确率的同时,缩短卷积神经网络训练时间。实验结果表明,该模型相比其他基于机器学习的入侵检测模型具有较高的准确率和较强的泛化能力,同时优化卷积神经网络训练方式,保证准确率的同时使神经网络训练时间减少了77%。Aiming at the problems of low accuracy,poor real-time performance and low generalization performance of existing network intrusion detection systems,we design a network intrusion detection model based on gain rate algorithm and convolutional neural network algorithm by using the advantages of deep learning,such as good classification performance and strong generalization ability.The gain rate was used to filter the data characteristics of data sets,which ensured the accuracy of intrusion detection and shortened the training time of convolutional neural network.The experimental results show that the intrusion detection model has higher accuracy and stronger generalization ability than other machine learning-based intrusion detection models,and optimizes the convolutional neural network training mode to ensure the accuracy and reduce the training time at 77%.

关 键 词：网络入侵检测 深度学习 卷积神经网络 增益率 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]