Research on Multimodality Face Antispoofing Model Based on Adversarial Attacks

Figure 5

An example of single-modality model attacking the RGB image, where (a) represents FGSM attack with the perturbation value of 0.01 and (b) represents DeepFool attack.