Abstract: To train robust malicious traffic identification models under noisy labeled datasets, a number of learning with noise labels approaches have been introduced, among which parallel training ...