Network anomaly traffic detection method based on ensemble learning and feature fusion

Aiming at the issues of incomplete feature information extraction and low detection accuracy in current network anomaly traffic detection methods, a Network Anomaly Traffic Detection Method Based on Ensemble Learning and Feature Fusion (NATDBELFF) is proposed. NATDBELFF integrates the spatial features extracted by Convolutional Neural Network (CNN) and Transformer, respectively, and utilizes Bidirectional Long Short-Term Memory (BiLSTM) to capture temporal features. It then ensembles the Decision Tree (DT) and Extreme Gradient Boosting (XGBoost) models to complete the recognition and classification of feature information. In this study, NATDBELFF is validated on the KDDCUP99, UNSW-NBV15, and NSL-KDD datasets. Experiments demonstrate that the features extracted by different models in the proposed method are distinctive and highly complementary. The fusion of these features results in a more complete representation. Furthermore, the integration of learning significantly enhances the overall classification performance.