ML Security PapersEnhancing Frequency Forgery Clues for Diffusion-Generated Image Detection
Daichi Zhang 1, Tong Zhang 1, Shiming Ge 2,3, Sabine Süsstrunk 1
Published on arXiv
2511.00429
Output Integrity Attack
OWASP ML Top 10 — ML09
Key Finding
F²C outperforms state-of-the-art diffusion-generated image detectors in generalization to unseen models and robustness to perturbations across multiple benchmark datasets.
F²C (Frequency Forgery Clue)
Novel technique introduced
Diffusion models have achieved remarkable success in image synthesis, but the generated high-quality images raise concerns about potential malicious use. Existing detectors often struggle to capture discriminative clues across different models and settings, limiting their generalization to unseen diffusion models and robustness to various perturbations. To address this issue, we observe that diffusion-generated images exhibit progressively larger differences from natural real images across low- to high-frequency bands. Based on this insight, we propose a simple yet effective representation by enhancing the Frequency Forgery Clue (F^2C) across all frequency bands. Specifically, we introduce a frequency-selective function which serves as a weighted filter to the Fourier spectrum, suppressing less discriminative bands while enhancing more informative ones. This approach, grounded in a comprehensive analysis of frequency-based differences between natural real and diffusion-generated images, enables general detection of images from unseen diffusion models and provides robust resilience to various perturbations. Extensive experiments on various diffusion-generated image datasets demonstrate that our method outperforms state-of-the-art detectors with superior generalization and robustness.
Key Contributions
- Empirical observation that diffusion-generated images exhibit progressively larger spectral differences from real images across low-to-high frequency bands
- Frequency Forgery Clue (F²C) representation using a learnable frequency-selective weighted filter over the Fourier spectrum to suppress uninformative bands and enhance discriminative ones
- Demonstrates superior generalization to unseen diffusion models and robustness to various perturbations compared to state-of-the-art detectors
🛡️ Threat Analysis
Directly addresses AI-generated content detection — specifically synthetic images from diffusion models — by proposing a novel frequency-selective detection technique. This is output integrity and content authenticity, a core ML09 concern.