基于增量式学习的网络攻击检测数学建模仿真

Mathematical Modeling and Simulation of Network Attack Detection Based on Incremental Learning

针对现有网络攻击检测数学模型,存在网络攻击检测率较低、网络攻击误报率较高、检测时效性较差等问题,构建一种基于增量式学习的网络攻击检测数学模型。首先利用拓扑几何学原理构建一个网络信息之间的拓扑结构关系,再利用该关系进行网络信息去噪,去噪后对数据做零均值化处理和对时间序列进行拟合。最后利用拟合得到的时间序列进行网络攻击判定实现检测,在此基础上利用增量式学习算法对网络信息数据做归一化处理,构建一个自回归数学模型,根据上述数学模型检测网络攻击。实验结果表明,所构建网络攻击检测数学模型具有高检测率、低误报率、高时效性的特点。

At present,the mathematical model of network attack detection has some problems,such as low detection rate,high false positive rate,and low timeliness.Therefore,a mathematical model of detecting the network attacks based on incremental learning was constructed.Firstly,a topological relation between network information was constructed based on the principle of topological geometry.On this basis,the network information was denoised.After the noise reduction,zero mean processing and time series fitting were performed on data.Finally,the fitting time series was used to determine the network attack.Moreover,the incremental learning algorithm was used to normalize the network information data and build an auto-regression mathematical model.According to the mathematical model,the network attack was detected.Simulation results show that the network attack detection mathematical model has high detection rate,low false positive rate and high timeliness.