Complete Defense Framework to Protect Deep Neural Networks against Adversarial Examples

Table 3

Performance of seven statistical detectors (%).

Positive examples	FGSM	R-FGSM	BIM	UAP	DeepFool	CW_UT	CW_T
Negative examples	Legitimate
TPR	99.8	100	99.8	99.4	85.8	81.9	81.1
FPR	0.5	0.2	0.9	0.6	26.0	25.5	26.6