Table 16 Model performance of different adversarial training methods.

Attack methods	Adversarial training methods	Evaluation metrics
Attack methods	Adversarial training methods	Accuracy	Precision	Recall	F1-score	Robust-accuracy
FGSM	Method one	0.9827	0.9823	0.9829	0.9826	0.9027
	Method two	0.9720	0.9715	0.9721	0.9716	0.8493
	Proposed methodology	0.9933	0.9931	0.9935	0.9933	0.9533
PGD	Method one	0.9813	0.9808	0.9817	0.9811	0.9000
	Method two	0.9773	0.9973	0.9779	0.9773	0.8507
	Proposed methodology	0.9907	0.9904	0.9906	0.9905	0.9467
BIM	Method one	0.9840	0.9838	0.9847	0.9840	0.9093
	Method two	0.9827	0.9824	0.9827	0.9825	0.8733
	Proposed methodology	0.9920	0.9916	0.9921	0.9918	0.9360

Quick links

Search