Publications by authors named "Huixin Yin"
Article Synopsis
- Traffic classification is crucial for detecting network anomalies and enhancing security, but current methods face challenges with feature design and data set limitations.
- The proposed BERT-based Time-Series Feature Network (TSFN) model incorporates both global and time-series features by using a BERT packet encoder and an LSTM module for improved accuracy.
- Testing the TSFN on the USTC-TFC dataset achieved an impressive F1 score of 99.50%, demonstrating the effectiveness of considering time-series features in malicious traffic classification.
