EUVIMEDEUVIMED entdecken
Crossref

Dynamic Attention Analysis for Backdoor Detection in Text-to-Image Diffusion Models

Zhongqi Wang, Jie Zhang, Shiguang Shan, Xilin Chen

IEEE Transactions on Pattern Analysis and Machine Intelligence · 2026-03 · 48 (3) · 3652-3665

DOI: 10.1109/tpami.2025.3644016 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026-03. Verfasst von Zhongqi Wang und 3 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 80 Literaturangaben. Gefördert von: Strategic Priority Research Program of the Chinese Academy of Sciences, Key Research and Development Program of Xinjiang Uyghur Autonomous Region, Beijing Nova Program.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Recent studies have revealed that text-to-image diffusion models are vulnerable to backdoor attacks, where attackers implant stealthy textual triggers to manipulate model outputs. Previous backdoor detection methods primarily focus on the static features of backdoor samples. However, a vital property of diffusion models is their inherent dynamism. This study introduces a novel backdoor detection perspective named Dynamic Attention Analysis (DAA), showing that these dynamic characteristics serve as better indicators for backdoor detection. Specifically, by examining the dynamic evolution of cross-attention maps, we observe that backdoor samples exhibit distinct feature evolution patterns at the $ token compared to benign samples. To quantify these dynamic anomalies, we first introduce DAA-I, which treats the tokens' attention maps as spatially independent and measures dynamic feature using the Frobenius norm. Furthermore, to better capture the interactions between attention maps and refine the feature, we propose a dynamical system-based approach, referred to as DAA-S. This model formulates the spatial correlations among attention maps using a graph-based state equation and we theoretically analyze the global asymptotic stability of this method. Extensive experiments across six representative backdoor attack scenarios demonstrate that our approach significantly surpasses existing detection methods, achieving an average F1 Score of 79.27% and an AUC of 86.27%.

Förderung

  • Strategic Priority Research Program of the Chinese Academy of Sciences
  • Key Research and Development Program of Xinjiang Uyghur Autonomous Region
  • Beijing Nova Program
  • Suzhou Frontier Technology Research Project
  • Youth Innovation Promotion Association of the Chinese Academy of Sciences

Zitieren

APA

Wang, Z., Zhang, J., Shan, S., & Chen, X. (2026). Dynamic Attention Analysis for Backdoor Detection in Text-to-Image Diffusion Models. IEEE Transactions on Pattern Analysis and Machine Intelligence, 48(3), 3652-3665. https://doi.org/10.1109/tpami.2025.3644016

Vancouver

Wang Z, Zhang J, Shan S, Chen X. Dynamic Attention Analysis for Backdoor Detection in Text-to-Image Diffusion Models. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2026;48(3):3652-3665. doi:10.1109/tpami.2025.3644016

BibTeX
@article{Wang2026,
  author = {Wang, Zhongqi and Zhang, Jie and Shan, Shiguang and Chen, Xilin},
  title = {Dynamic Attention Analysis for Backdoor Detection in Text-to-Image Diffusion Models},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year = {2026},
  volume = {48},
  number = {3},
  pages = {3652--3665},
  doi = {10.1109/tpami.2025.3644016}
}
Kostenlos starten →