基于深度卷积神经网络的无线通信网络异常攻击检测  

Abnormal Attack Detection in Wireless Communication Network Based on Deep Convolutional Neural Network

作　　者：谭伦荣 王辉 TAN Lunrong;WANG Hui(School of Computer Science,Huainan Normal University,HuainanAnhui 232001,China)

机构地区：[1]淮南师范学院计算机学院,安徽淮南232001

出　　处：《重庆科技学院学报（自然科学版）》2022年第6期60-64,共5页Journal of Chongqing University of Science and Technology：Natural Sciences Edition

基　　金：安徽省教育厅科研项目“基于FBM的混合学习交互模型的研究”(KJ2021A0972)。

摘　　要：为了解决无线通信网络异常攻击检测的准确率和查全率低等问题,提出一种基于深度卷积神经网络的无线通信网络异常攻击检测方法。设计卷积神经网络,提取并处理无线通信网络日志特征;构建深度卷积神经网络模型,将日志特征作为模型输入值,引入非线性激活函数进行模型学习;确定各层级参数,构建异常攻击检测框架,设置检测步骤,以实现无线通信网络异常攻击检测。实验结果表明,该方法的分类精度高、训练损失小,准确率、查全率、误报率平均值分别为99.96%、99.94%和0.01%,具有一定的可行性和较高的应用价值。To solve the problem of low detection accuracy and recall rate of abnormal attacks in wireless communication networks,a new anomaly attack detection method based on deep convolution neural network is proposed.Convolutional neural network isdesignedto extract log features of wireless communication network.A deep convolution neural network model isbuilt,and the log characteristics are taken as the input value of the model.Non-linear activation functions are introduced to complete the learning process of the model.The parameters of each level aredetermined to build an anomaly attack detection framework,and detection steps are set torealize detection of abnormal attacks on wireless communication networks.The experimental results show that the detection method has high classification accuracy,small training loss,and it performs well in the accuracy,recall,and false alarm rate of network anomaly attack detection,with the average values of 99.96%,99.94%,and 0.01%,respectively,which verifies the feasibility and higher application value of the method.

关 键 词：卷积神经网络 无线通信网络 异常攻击 网络日志 二值化处理 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]