A Network Security Situation Assessment Method Based on Multi-attention Mechanism and HHO-ResNeXt

Abstract

Because the traditional convolutional neural network (CNN) cannot obtain the importance of each channel and its receptive field is limited, it is difficult to deal with the increasingly complex network environment. Aiming at the shortcomings, this paper combines the Contextual Transformer (COT) block and the Efficient Channel Attention (ECA) module with ResNeXt and uses Harris Hawks Optimization (HHO) algorithm to choose the most suitable hyperparameters to improve the model performance. This model enables the rich contexts among neighbor keys to be fully exploited and can obtain the importance of each channel to improve the weight of the useful channel and suppress the less useful channel, effectively making up for the shortcomings of the traditional convolutional neural network. The experiments on a network security dataset show that the model is superior to other models in the network security situation assessment effect and its comprehensive performance in accuracy, precision and F-score.