Implementation and evaluation for Deep Learning Project 3 (Spring 2025, NYU Tandon). We attack a pretrained ResNet-34 model using ℓ∞-bounded adversarial perturbations, including FGSM, PGD, Momentum PGD, and Patch PGD, and assess transferability to DenseNet-121.
machine-learning jupyter-notebook deeplearning nyu fgsm resnet-34 pgd-adversarial-attacks fgsm-attack patch-pgd densene
-
Updated
May 22, 2025 - Jupyter Notebook