基于改进差分进化算法的网络入侵检测方法

A network intrusion detection method based on improved differential evolution algorithms

为提高网络入侵检测模型的准确率或降低误报率,解决传统神经网络模型建模时间过长等问题,提出基于改进的差分进化算法的网络入侵检测模型GSDEGRU。首先使用SMOTE技术对数据进行平衡约简处理,其次将引力搜索算子加入到差分进化算法中优化GRU模型超参数,提高其全局寻优能力。最后通过GSDEGRU神经网络算法对网络入侵数据进行识别并分类。实验结论表明,GSDEGRU神经网络模型准确率比传统的GRU模型高出1.9%,建模时间缩短了322 s;准确率、精确率和召回率好于RNN、DNN等深度学习模型。

To improve the accuracy or reduce the false alarm rate of network intrusion detection model,and to solve the problems such as the long modeling time of traditional neural network models,a network intrusion detection model GSDEGRU based on improved differential evolution algorithms is proposed. Firstly,the data is balanced and reduced using SMOTE technique. Secondly,the gravitational search operator is added to the differential evolution algorithm to optimize the GRU model hyperparameters and improve its global optimization capability. Finally,the network intrusion data are identified and classified by the GSDEGRU neural network algorithm. The experimental results show that the accuracy of the GSDEGRU neural network model is1.9% higher than the traditional GRU model,and the modeling time is shortened by 322 s. The accuracy,precision,and recall of the new model are better than the deep learning models such as RNN and DNN.