D$^3$: Scaling Up Deepfake Detection by Learning from Discrepancy

arXiv:2404.04584v1 Announce Type: new
Abstract: The boom of Generative AI brings opportunities entangled with risks and concerns. In this work, we seek a step toward a universal deepfake detection system with better generalization and robustness, to accommodate the responsible deployment of diverse image generative models. We do so by first scaling up the existing detection task setup from the one-generator to multiple-generators in training, during which we disclose two challenges presented in prior methodological designs. Specifically, we reveal that the …

abstract arxiv boom concerns cs.cv deepfake deployment detection diverse generative generative models image opportunities responsible risks robustness scaling scaling up type universal work