Network traffic classification method based on attention-CNN-BiLSTM

With the wide application of artificial intelligence, cloud computing and Internet of things, we have entered a new era of intelligent interconnection. However, network security threats and risks are also increasing. At present, it is urgent to solve how to effectively combat network attacks and ensure network security. Network traffic analysis plays a key role in intrusion detection. By accurately classifying network traffic, network intrusion can be more targeted. However, due to the imbalance of categories in the data samples, the recognition accuracy of minority classes is low. In order to cope with this challenge, an improved network traffic analysis method based on Attention-CNN-Bi LSTM is proposed. This method first balances the data set, then uses the feature selection strategy to eliminate unnecessary features, and uses the combined neural network model for training. The experimental simulation of this method shows that the AUC value of the minority class exceeds 0.85 on both data sets, which shows that the method has good generalization performance for the minority class and solves the problem of low accuracy of the minority class in the unbalanced data set.