基于残差卷积神经网络的网络安全态势感知方法  

A situational⁃aware method for network security based on residual convolutional neural network

作　　者：李立 Li Li(Ezhou Branch of China United Network Communications Co.,Ltd.,Ezhou 436000,China)

机构地区：[1]中国联合网络通信有限公司鄂州市分公司,鄂州436000

出　　处：《现代计算机》2024年第9期56-60,共5页Modern Computer

摘　　要：由于影响网络安全态势的因素具有多元化的特征,网络安全态势的观察值与预测值也是不断变化的。这种波动导致传统的神经网络在对其进行感知时,对应的收敛误差难以控制。提出基于残差卷积神经网络的网络安全态势感知方法研究。分别从网络自身和攻击状态两个角度,对网络安全态势影响因素进行量化分析;再利用卷积核的权重系数对输入神经网络的整体状态参数进行加权平均,提取各网络安全态势影响因素状态。引入残差损失参数对残差卷积神经网络的池化结果进行约束,输出最终的网络安全态势值。在测试结果中:收敛误差值面对不同类型的网络流量和攻击手段表现出了较高的稳定性,且始终处于较低水平,收敛误差最大值仅为0.0345。Due to the diversified factors affecting the network security situation,the observed value and forecast value of the network security situation are also constantly changing.This fluctuation leads to the corresponding convergence error of traditional neural networks.The research on network security situational awareness based on residual convolutional neural network.From the perspective of the network itself and the attack state,the influencing factors of the network security situation,and then the weighted average of the overall state parameters of the input neural network to extract the state of the network security situation.The residual loss parameter is introduced to restrict the pooling results of the residual convolution neural network,and the final network security situation value is output.In the test results:the convergence error value shows high stability against different types of network traf-fic and attack means,and is always at a low level,and the maximum convergence error value is only 0.0345.

关 键 词：残差卷积神经网络 网络安全态势感知 影响因素 量化分析 加权平均 残差损失参数 收敛误差值 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]