Computer Network Security Situation Awareness using One Dimensional Convolutional Neural Network
Moulong Liu · 2024
The rapid development of computer network repressed the existing security system strength and affects data confidentiality. However, the existing classifier has limitation such as overfitting due to ineffective training. Therefore, One Dimensional Convolutional Neural Network (1D-CNN) is proposed a computer network security situation awareness. The 1D-CNN effectively capture spatial patterns and reduces computation cost. Additionally, it required less parameters for enhancing performance and minimizing overfitting issues. The Information Gain (IG) technique is used for selecting features which attains constant set of selected features and prevent the overfitting issues. The UNSW-NB15 dataset is used and it is preprocessed by min-max normalization which rescales the data into the range of 0 and 1 thereby leading quick convergence. The accuracy, f1-score, recall and precision are taken as metrics to estimate the 1D-CNN performance. The 1D-CNN attained 98.64%, 95.79%, 95.13% and 96.37% of accuracy, f1-score, recall and precision for UNSW-NB15 dataset compared to Harris Hawks Optimization (HHO)-ResNeXt.