Intelligent gateway training and flow detection method and device

The invention discloses an intelligent gateway training and traffic detection method and device, and belongs to the technical field of malicious traffic detection. Comprising the steps of preprocessing training data, filtering out unencrypted data packets in the training data, and obtaining formatted encrypted text data; the formatted encrypted text data is distributed in a balanced mode through an enhanced domain generalization algorithm, and data service groups are formed; performing data feature extraction on the data service groups by using a natural language processing model, and converting qualitative data in the data service groups into quantitative data; capturing an optimal data feature subset based on an incremental random walk feature selection algorithm; and performing optimization training on the data feature subset to obtain the encrypted malicious traffic detection intelligent gateway. According to the method, the influence of data imbalance in the security field on the performance of the model