Semantic-Preserving Multi-Object Coexistence: A Backdoor Attack on Text-to-Image Diffusion Models
Mathematics · 28.05.2026 · 14 (11) · 1874
DOI: 10.3390/math14111874 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mathematics, online erschienen am 28.05.2026. Verfasst von Zhoufan Yang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 52 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Hainan Provincial Natural Science Foundation of China, project of Hunan Provincial Social Science Achievement Review Committee of China.
Abstract (Originalsprache)
Text-to-image (T2I) diffusion models have become popular in computer vision, but they remain vulnerable to backdoor attacks. Existing methods typically trigger a fixed image regardless of user input, causing severe semantic inconsistency between the generated image and the original prompt. This makes the attack easily detectable by machines as it would lack visual stealth. To overcome this challenge, we propose MultiAttack, a novel semantic-preserving multi-object coexistence backdoor attack for T2I diffusion models, which retains prompt-described objects while injecting malicious targets. First, we propose a semantic-preserving data poisoning strategy to build a latent mapping, which maps the trigger into a composite semantic space while retaining the original prompt context. Second, we design a backdoor enhancement mechanism to embed the spatial orthogonality between malicious and benign objects into model weights as a conditional response, which strengthens the model’s ability to generate stable malicious outputs without requiring additional inference. Results on Stable Diffusion show that compared tostate-of-the-art baselines, MultiAttack increases attack success rate by 13.1% and visual stealth (defined as the success rate of co-generating both prompt-described and target objects) by 12.6%, with an FID increase of less than 1.2 and a CLIP score decrease of under 1 compared to clean models.
Förderung
- National Natural Science Foundation of China
- Hainan Provincial Natural Science Foundation of China
- project of Hunan Provincial Social Science Achievement Review Committee of China
Zitieren
APA
Yang, Z., Ning, H., Pan, Y., Liao, J., & Zhang, S. (2026). Semantic-Preserving Multi-Object Coexistence: A Backdoor Attack on Text-to-Image Diffusion Models. Mathematics, 14(11), 1874. https://doi.org/10.3390/math14111874
Vancouver
Yang Z, Ning H, Pan Y, Liao J, Zhang S. Semantic-Preserving Multi-Object Coexistence: A Backdoor Attack on Text-to-Image Diffusion Models. Mathematics. 2026;14(11):1874. doi:10.3390/math14111874
BibTeX
@article{Yang2026,
author = {Yang, Zhoufan and Ning, Honghui and Pan, Yimeng and Liao, Junguo and Zhang, Shaobo},
title = {Semantic-Preserving Multi-Object Coexistence: A Backdoor Attack on Text-to-Image Diffusion Models},
journal = {Mathematics},
year = {2026},
volume = {14},
number = {11},
pages = {1874},
doi = {10.3390/math14111874}
}
EUVIMED