Network traffic classification method based on sample playback and terminal

By adopting the method of sample playback and knowledge distillation in network traffic classification, the sample set is constructed and enhanced and hyperparameter balance is introduced, the problem of poor adaptability and high resource consumption of deep learning models when traffic changes is solved, and a higher level is achieved classification accuracy and stability.

CN119202873AInactive Publication Date: 2024-12-27STATE GRID FUJIAN ELECTRIC POWER RES INST +1
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202410877844.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-02
Publication Date
2024-12-27
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure CN119202873A_ABST
    Figure CN119202873A_ABST
Patent Text Reader

Abstract

The invention discloses a network traffic classification method and terminal based on sample playback, and the method comprises the steps: training a traffic classification model based on network traffic, constructing a class typical sample set and a class sampling sample set based on the trained traffic classification model and the network traffic, carrying out the data enhancement of the class typical sample set and the class sampling sample set, and carrying out the classification of the class typical sample set and the class sampling sample set. And training the trained traffic classification model by using a knowledge distillation strategy based on the enhanced category typical sample set, the category sampling sample set and new network traffic to obtain a new traffic classification model, and repeating the sample set construction and model training steps until the end of the current preset incremental learning cycle. In this way, deviation of the model to new data is slowed down, even if the network flow is continuously updated and changed, it can be ensured that the flow classification model always keeps the optimal flow classification capacity, flow classification can be more accurately achieved through the flow classification model, and therefore the flow classification accuracy is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Online adaptive intrusion detection method based on incremental learning

    CN115130102A

  • Network traffic classification method based on long-tail incremental learning

    CN117614678A

  • Deep-neural-network-based class-incremental learning method for mobile phone radiation source spectrogram

    WO2024119422A1