OpenCBD: A Network-Encrypted Unknown Traffic Identification Scheme Based on Open-Set Recognition

Table 6

When the number of encoders in the ensemble model of OpenCBD is 4, 8, and 12, accuracy, precision, recall, and F1-score of the 2-class classification tasks (identifying known and unknown classes) vary as the number of known classes is randomly selected in the first training stage.

	Accuracy			Precision			Recall			F1-score
4	8	12	4	8	12	4	8	12	4	8	12
3	0.5607	0.603	0.613	0.714	0.6853	0.6951	0.4775	0.6562	0.6612	0.5722	0.6704	0.6777
4	0.6615	0.6853	0.6869	0.7337	0.7459	0.7484	0.7062	0.7412	0.74	0.7197	0.7435	0.7441
5	0.6269	0.73	0.7238	0.7215	0.765	0.778	0.6412	0.81	0.7712	0.67902	0.7868	0.7746
6	0.693	0.7307	0.7207	0.775	0.7952	0.7841	0.7062	0.7575	0.7537	0.739	0.7759	0.7686
7	0.7238	0.7484	—	0.7766	0.7832	—	0.7737	0.8175	—	0.7752	0.7999	—