基于原子搜索优化深度神经网络的网络安全态势预测

Network Security Situation Prediction Based on Atom Search Optimized Deep Neural Network

为提高网络安全态势预测准确度,将深度卷积神经网络(CNN)运用于安全态势预测,并借助原子搜索算法改进深度卷积神经网络,以提高其在网络安全态势预测方面的适应度。首先,提取网络样本流量特征并完成初始化,接着建立深度CNN网络攻击检测模型,并采用原子搜索优化(ASO)算法对CNN网络参数进行优化求解。通过原子适应度、质量及加速度的计算,不断更新原子的速度和位置,以获得最高适应度的CNN网络参数原子个体。然后采用最优参数进行CNN网络攻击类型检测训练,确定网络攻击类型。最后根据攻击类型权重和主机权重计算网络安全态势预测值。实验证明,在合理设置主机权重的情况下,通过ASO-CNN算法获得的网络安全态势预测值精度高,且稳定性强。

In order to improve the accuracy of network security situation prediction,the deep convolution neural network(CNN)was applied to security situation prediction,and the atomic search algorithm was used to improve the depth convolution neural network to improve its adaptability in network security situation prediction.first,extracted the characteristics of network sample traffic and complete initialization,then established a deep CNN network attack detection model,and optimized the CNN network parameters with atomic search optimization(ASO)algorithm.Through the calculation of atomic fitness,mass and acceleration.The speed and position of the new atom to obtain the highest fitness CNN network parameter atomic individual,and then used the optimal parameters for CNN network attack type detection training to determine the type of network attack,and finally calculated the network security state according to the attack type weight and host weight potential prediction value.The experiment proved that the network security situation prediction value obtained through ASO-CNN algorithm was of high accuracy and stability when the host weight was reasonably set.