UniNDM: A Unified Noise-Driven Detection and Mitigation Framework Against Sexual Content in Text-to-Image Generation
IEEE Transactions on Pattern Analysis and Machine Intelligence · 2026
DOI: 10.1109/tpami.2026.3713771 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026. Verfasst von Yao Huang und 10 weiteren Autor:innen. Hauptthemen laut MEDLINE: Image Processing, Computer-Assisted. Die Arbeit nennt 55 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities.
Abstract (Originalsprache)
Despite the impressive generative capabilities of text-to-image (T2I) diffusion models, they remain vulnerable to implicit sexual prompts, where subtle cues disguised as benign terms or adversarial tokens unexpectedly generate the inappropriate content due to model biases or latent correlations in training data. Existing safety mechanisms face fundamental limitations: detection methods primarily identify explicit content and fail to capture implicit malicious intent, while mitigation approaches rely on static negative prompts inadequate for diverse implicit scenarios. To address these challenges, we propose UniNDM, a unified noise-driven framework that rethinks safety mechanisms through the lens of noise dynamics in diffusion processes. Our key insight is that early-stage predicted noise exhibits inherent separability between normal and sexually explicit content, which we theoretically demonstrates quadratically increasing semantic concentration with timestep. Leveraging this property, we develop a lightweight noise-based detector achieving superior accuracy with virtually no computational overhead. For mitigation, we introduce noise-enhanced adaptive negative guidance: dynamically generating context-specific negative prompts via large language models to handle diverse implicit content, while optimizing initial noise by suppressing attention concentration on explicit tokens to provide comprehensive protection. Besides the U-Net-based diffusion models, we further extend our framework to emerging Diffusion Transformer architectures through region-constrained semantic guidance tailored for their unified multimodal attention. Comprehensive experiments across U-Net models (SDv1.4, v1.5, v2.1, XL) and DiT models (SDv3) on both natural and adversarial datasets demonstrate substantial improvements over state-of-the-art methods, including SLD, UCE, Safree, etc.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Förderung
- National Natural Science Foundation of China
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Huang, Y., Sun, Y., Chen, H., Zhang, R., Ruan, S., Duan, R., Yuan, M., Dong, Y., Xue, H., Cao, X., & Wei, X. (2026). UniNDM: A Unified Noise-Driven Detection and Mitigation Framework Against Sexual Content in Text-to-Image Generation. IEEE Transactions on Pattern Analysis and Machine Intelligence. https://doi.org/10.1109/tpami.2026.3713771
Vancouver
Huang Y, Sun Y, Chen H, Zhang R, Ruan S, Duan R, et al. UniNDM: A Unified Noise-Driven Detection and Mitigation Framework Against Sexual Content in Text-to-Image Generation. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2026. doi:10.1109/tpami.2026.3713771
BibTeX
@article{Huang2026,
author = {Huang, Yao and Sun, Yitong and Chen, Huanran and Zhang, Ruochen and Ruan, Shouwei and Duan, Ranjie and Yuan, Maoxun and Dong, Yinpeng and Xue, Hui and Cao, Xiaochun and Wei, Xingxing},
title = {UniNDM: A Unified Noise-Driven Detection and Mitigation Framework Against Sexual Content in Text-to-Image Generation},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
year = {2026},
doi = {10.1109/tpami.2026.3713771}
}
EUVIMED