Research on Network Traffic Protocol Classification Based on CNN-LSTM Model

Network traffic analysis is essential for network security and performance optimization, yet classifying network traffic protocols remains a challenge. This study improves the classification and prediction of unknown network traffic protocols. By collecting and analyzing extensive traffic data we examine the correlation between traffic characteristics and protocol types.We introduce a CNN-LSTM model that integrates Convolutional Neural Networks (CNN) for local perception and weight sharing, and Long Short-Term Memory (LSTM) networks for temporal sequence modeling, which improves the accuracy of protocol classification. Experiments show that the CNN-LSTM model outperforms other models in terms of accuracy and F1 score. With feature selection, the accuracy reaches 0.981; while with raw features, both accuracy and F1 score reach 0.956. In contrast, standalone LSTM and CNN models show weaker performance and are more sensitive to changes in the number of features.This study validates the effectiveness of the CNN-LSTM model for network traffic protocol classification, and provides insights for future research. Future studies may explore ways to optimize the model structure and feature processing to cope with more complex network environment and traffic data.