网络多入侵行为识别的数学建模与仿真  

Mathematical modeling and Simulation of network multi intrusion behavior identification

作　　者：潘宝柱 魏文英 昝立荣 张秀珍 PAN Bao-zhu;WEI Wen-ying;ZAN Li-rong;ZHANG Xiu-zhen(Hebei Polytechnic Institute,Shijiazhuang Hebei 050091,china;School of Mathematics&Statistics Shanxi Datong University,Datong Shanxi 037009,China)

机构地区：[1]河北工程技术学院,河北石家庄050091 [2]山西大同大学数学与统计学院,山西大同037009

出　　处：《计算机仿真》2022年第2期357-360,446,共5页Computer Simulation

摘　　要：当前的网络多入侵行为识别方法无法计算入侵信任度,导致传统方法下网络流量数据振幅识别结果波动不大,识别结果存在失真问题。为解决上述问题,基于信任度计算设计多入侵识别数学建模仿真方法。通过构建动静结合的信任度模型,计算网络节点多入侵信任度,并输入至基于深度学习神经网络多入侵识别数学模型中。利用自适应特征映射和深度学习神经网络算法,通过确定各网络节点的权值实现网络多入侵特征匹配。并通过深度学习神经网络调整模型的误差,实现网络多入侵识别。实验结果表明,上述方法通过引入动态信任度反映节点的信任度变化,对网络多入侵识别具有积极作用。所提方法识别输出的网络流量数据振幅变化显著,且识别结果与实际结果相同并具有稳定区间,有效提高了网络多入侵识别能力,合理规避网络风险。Currently, the network multi-intrusion behavior identification method without calculating intrusion trust leads to large fluctuation of network traffic data amplitude identification results and low accuracy. Therefore, we report a mathematical modeling and simulation method of multi-intrusion recognition based on trust computing. The dynamic and static trust model was built to calculate the multi-intrusion trust of network nodes. Then, the trust results were input into the mathematical model of multi-intrusion recognition based on deep learning neural network. According to adaptive feature mapping and deep learning neural network algorithm, the weights of each network node were obtained to realize network multi-intrusion feature matching. Based on the error of the model adjusted by the deep learning neural network, the network multi-intrusion recognition was completed. The experimental results show that the amplitude of the network traffic data identified by this method changes significantly, the identification results are consistent with the actual results, and the interval is stable.

关 键 词：信任度计算 多入侵识别 建模仿真 自适应特征映射 深度学习神经网络算法 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]