Complete Defense Framework to Protect Deep Neural Networks against Adversarial Examples

Table 4

Performance of the single statistical detector (%).

Positive examples	FGSM	R-FGSM	BIM	UAP	Overall
TPR	99.9	100	100	98.3	99.6
Negative examples	DeepFool	CW_UT	CW_T	Legitimate	Overall
FPR	0.6	0.6	0.6	0.7	0.6