Robust Face Verification via Disentangled Representations

Abstract

We introduce a robust algorithm for face verification, i.e., deciding whether twoimages are of the same person or not. Our approach is a novel take on the idea ofusing deep generative networks for adversarial robustness. We use the generativemodel during training as an online augmentation method instead of a test-timepurifier that removes adversarial noise. Our architecture uses a contrastive loss termand a disentangled generative model to sample negative pairs. Instead of randomlypairing two real images, we pair an image with its class-modified counterpart whilekeeping its content (pose, head tilt, hair, etc.) intact. This enables us to efficientlysample hard negative pairs for the contrastive loss. We experimentally show that, when coupled with adversarial training, the proposed scheme converges with aweak inner solver and has a higher clean and robust accuracy than state-of-the-art-methods when evaluated against white-box physical attacks.

arXiv commentPreprint

References in corpus (13)

• Adam: A Method for Stochastic Optimization
• Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
• Improved Regularization of Convolutional Neural Networks with Cutout
• Theoretically Principled Trade-off between Robustness and Accuracy
• Fast is better than free: Revisiting adversarial training
• Towards the first adversarially robust neural network model on MNIST
• Improving Robustness Without Sacrificing Accuracy with Patch Gaussian Augmentation
• Attacks Which Do Not Kill Training Make Adversarial Learning Stronger
• Certified Defenses for Adversarial Patches
• Adversarial Manipulation of Deep Representations
• Disentangling Disentanglement in Variational Autoencoders
• Curriculum Adversarial Training
• GAT: Generative Adversarial Training for Adversarial Example Detection and Robust Classification