Deepfake Detection without Deepfakes: Generalization via Synthetic Frequency Patterns Injection

arXiv:2403.13479v1 Announce Type: new
Abstract: Deepfake detectors are typically trained on large sets of pristine and generated images, resulting in limited generalization capacity; they excel at identifying deepfakes created through methods encountered during training but struggle with those generated by unknown techniques. This paper introduces a learning approach aimed at significantly enhancing the generalization capabilities of deepfake detectors. Our method takes inspiration from the unique "fingerprints" that image generation processes consistently introduce into the frequency domain. These fingerprints manifest as …

abstract arxiv capacity cs.ai cs.cv deepfake deepfake detectors deepfakes detection excel generated images paper patterns struggle synthetic through training type via