Harnessing the Vulnerability of Latent Layers in Adversarially Trained Models
Neural networks are vulnerable to adversarial attacks – small visually imperceptible crafted noise which when added to the input …
PDF