A linear discriminant analysis-based algorithm for identifying anomalous traffic in large-scale networks

Abstract

Abstract To protect network security, this paper develops a large-scale network anomalous traffic identification algorithm that utilizes the linear discriminant analysis method to intercept network anomalous traffic. Firstly, the classification of large-scale network anomalous traffic is explored, and the SSAE algorithm is combined with the feature selection of large-scale network traffic on the basis of network flow feature extraction. Secondly, data dimensionality reduction of network anomalous traffic using linear discriminant analysis and feature selection of large-scale network traffic based on SSAE to identify network anomalous traffic. Finally, the CICIDS2017 dataset and the NSL-KDD dataset are used to experimentally analyze the effect and performance of feature selection and anomaly identification algorithms. The results show that the classification accuracy of the feature selection algorithm is 0.989, the 10-dimensional optimal features selected are (12,6,5,38,29,3,33,35,36,40), and the recognition result is 0.803 for normal network traffic and 0.197 for anomalous traffic, with an overall recognition error of 0.003, and a performance of more than 0.988.