AI Paper Atlas

Research map / Image generation / GANs

Artificial Fingerprinting for Generative Models: Rooting Deepfake Attribution in Training Data

Ning Yu, Vladislav Skripniuk, Sahar Abdelnabi, Mario Fritz

ICCV 2021 · Image generation › GANs · approach: GAN latent optimization and evaluation · 221 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
embed fingerprints in training data instead of jointly training the GAN to emit themProGAN · CelebAFID: 60.28 -> 14.38Improved
use deep-learning steganography instead of conventional JPEG-based steganographyStyleGAN2 · CelebAbitwise accuracy: 0.51 -> 0.99Improved
embed fingerprints in training data instead of jointly training the GAN to emit themProGAN · CelebAbitwise accuracy: 0.93 -> 0.98Improved
fingerprint the training dataProGAN · LSUN CatFID: 45.22 -> 48.97Hurt
fingerprint the training dataStyleGAN2 · CelebAFID: 6.41 -> 6.23No clear effect
fingerprint the training dataBigGAN · CIFAR-10FID: 6.25 -> 6.80Hurt
add Gaussian noise (std 0.1) to generated imagesfingerprinted ProGAN · CelebAbitwise accuracy: 0.99 -> 0.77Hurt
apply model weight quantization (precision 10^0)fingerprinted ProGAN · CelebAbitwise accuracy: 0.99 -> 0.64Hurt
use artificial fingerprints instead of a CNN deepfake detectorProGAN · CelebA deepfake detectiondetection accuracy: 0.508 -> 1.000Improved
use artificial fingerprints instead of a CNN classifier for attributionProGAN · CelebA deepfake attributionattribution accuracy: 0.235 -> 1.000Improved
give the attacker access to encoder and decoder with fingerprint codes 10 bits differentfingerprint secrecy · ATS attackattacker classification accuracy: 0.503 -> 0.504No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy