通过分阶段训练提升GAN生成人脸的检测精度，达到99%以上。
IRIS-GAN: Staged Specialist Detection of Deepfake Faces
- 分阶段引入更复杂的GAN生成器进行训练，逐步增强检测能力。
- 在多个GAN家族上检测准确率超99%，真实人脸识别率达98.9%。
- 可识别非GAN类深度伪造，适合高精度人脸伪造检测场景。
我们提出IRIS-GAN，一种针对跨生成器迁移下合成人脸图像的专用检测器。不同于通用合成图像检测，本文聚焦于生成对抗网络（GAN）生成的人脸——当前主流的深度伪造内容形式。通过分阶段暴露于日益复杂的GAN家族并保留早期生成器，模型最终在所考虑的GAN家族中实现超过99%的伪造检测率，并在外部真实人脸数据集上达到98.9%的分类准确率。Grad-CAM分析揭示了可量化的、与生成器相关的空间响应模式，这些模式对仅使用热图的二级分类器仍具信息价值。在扩散模型生成的人脸上进行的跨家族测试表明，IRIS-GAN是专用检测器，具备一定识别非GAN类深度伪造的能力。这些结果证实分阶段训练是构建鲁棒GAN人脸取证的有效策略。
原文摘要 · Abstract (English)
We introduce IRIS-GAN, a specialist forensic detector for synthetic face images under cross-generator shift. Rather than addressing universal synthetic-image detection, we focus on faces generated by generative adversarial networks (GANs), which are state-of-the-art in deepfake content, and train the detector through staged exposure to increasingly demanding GAN families while retaining earlier generators. The final model reaches fake-detection rates above 99% across the GAN families considered and classifies an external real-face dataset with 98.9% accuracy. Grad-CAM analysis further reveals measurable generator-dependent spatial response patterns, which remain informative for a secondary heatmap-only classifier. Out-of-family tests on diffusion-generated faces confirm that IRIS-GAN is a specialist detector, with some capability to reach non-GAN deepfakes. These results establish staged training as an effective strategy for robust GAN-face forensics.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。