一种基于SA-SOA-BP神经网络的网络安全态势预测算法

Network Security Situation Prediction Algorithm Based on SA-SOA-BP Neural Network

网络安全态势预测能够依据已有的网络安全数据预测网络未来的安全状况及其变化趋势,为安全策略的选取提供指导,从而增强网络防御的主动性,尽可能地降低危害.然而现有的网络安全态势预测方法的精准度和收敛性还不理想.为了提高网络安全态势预测的准确性,提出了一种将模拟退火算法(SA)引入人群搜索算法(SOA)优化BP神经网络的网络安全态势预测方法.该算法利用人群搜索算法特有的利己、利他、预动和不确定推理四大行为特征确定搜索策略,找到最佳适应度个体,获取最优权值和阈值,然后再对BP神经网络的随机初始阈值和权值进行赋值,经过神经网络训练后得到预测值.针对人群搜索算法在搜索后期易陷入局部最优和收敛缓慢等问题,又将模拟退火算法引入人群搜索算法,根据它的Metropolis准则以一定的概率接受恶解,避免了算法陷入局部最优的陷阱,提高了该算法的全局搜索能力.与其它基于改进BP神经网络的预测算法进行对比的实验表明,该优化算法准确性更高,稳定性更强,收敛效果更好.

Netw ork security situation prediction can predict the future security situation and its changing trend based on the existing netw ork security data,and provide guidance for the selection of security strategies,so as to enhance the initiative of netw ork defense and reduce the harm as much as possible.How ever,the accuracy and convergence of existing netw ork security situation prediction methods are not ideal.In order to improve the accuracy and convergence of netw ork security situation prediction,a netw ork security situation prediction method based on improved BP neural netw ork optimized by introducing simulated annealing(SA)algorithm into seeker optimization algorithm(SOA)is proposed.This algorithm uses the four behavioral characteristics of seeker optimization algorithm:self-interest,altruism,pre-action and uncertain reasoning to determine the search strategy,find the best fitness individual,obtain the best w eight and threshold value,then assign them to the random initial threshold and w eight value of BP neural netw ork,and get the prediction value after the training of neural netw ork.But the seeker optimization algorithm is prone to fall into local optimization and slow convergence in the later stage of the search.In order to improve its deficiencies,the simulated annealing algorithm is introduced into the seeker optimization algorithm.According to the M etropolis criterion of the simulated annealing algorithm,the bad solution is accepted w ith a certain probability,w hich avoids the algorithm falling into the trap of local optimization and improves the global search ability of the algorithm.Compared w ith other prediction algorithms based on optimized BP neural netw ork,the experimental results show that this optimized algorithm has higher accuracy,stronger stability and better convergence.