AEGuard: Image Feature-Based Independent Adversarial Example Detection Model

Table 1

Adversarial attack block mechanism comparison.

	Adversarial noise reduction [9]	Robust model [10]	Statistics-based detection [11]	Proposed system
Adversarial example detection accuracy	Low	High	Low	High
Classification accuracy decrease	High	Low	No	No
Input image modification	O	X	X	X
Independence	O	X	X	O
Versatility	O	X	O	O