Identifying IoT Devices Based on Spatial and Temporal Features from Network Traffic

Table 9

Performance comparison of different model combinations.

Method		Accuracy	Precision	Recall	F1-score
CNN	UNSW dataset	0.9464	0.8919	0.8858	0.8872
	Laboratory dataset	0.9302	0.9026	0.8774	0.8877
FGAN + CNN	UNSW dataset	0.9517	0.8990	0.8963	0.8950
	Laboratory dataset	0.9463	0.9081	0.8925	0.8989
BiLSTM	UNSW dataset	0.9465	0.8897	0.8800	0.8812
	Laboratory dataset	0.9382	0.8875	0.8581	0.8664
FGAN + BiLSTM	UNSW dataset	0.9541	0.9059	0.8882	0.8906
	Laboratory dataset	0.9478	0.8900	0.8671	0.8754
CNN + BiLSTM	UNSW dataset	0.9865	0.9325	0.8865	0.9039
	Laboratory dataset	0.9524	0.9033	0.8752	0.8850
FGAN + CNN + BiLSTM (CBBI)	UNSW dataset	0.9983	0.9958	0.9914	0.9935
	Laboratory dataset	0.9726	0.9651	0.9652	0.9649