EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

ADRD: Detecting diffusion-generated images via adversarial perturbation induced reconstruction discrepancy

Yi Zhou iD, Xiangwei Hu, Jun Tong, Changting Lin iD, Mei Sun, Shuangxi Chen iD

PLOS One · 14.07.2026 · 21 (7) · e0350655

DOI: 10.1371/journal.pone.0350655 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS One, online erschienen am 14.07.2026. Verfasst von Yi Zhou und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Internet, Jiaxing Vocational and Technical College, Jiaxing, Zhejiang, China. Hauptthemen laut MEDLINE: Image Processing, Computer-Assisted. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Education Department of Zhejiang Province, Pioneer and Leading Goose R&D Program of Zhejiang, Scientific Research Foundation of Jiaxing Vocational and Technical College.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid advancement of diffusion models has raised concerns about their misuse in generating deceptive visual content, motivating growing interest in AI-generated image detection. Many existing detection methods rely on image semantic features, but modern diffusion models are optimized to closely match the semantic structure of real images, reducing the effectiveness of semantic-based detection. An alternative line of work exploits differences revealed through diffusion reconstruction; however, most existing approaches treat reconstruction error as a static and passive metric, which can be sensitive to generators or post-processing, thereby limiting robustness. In this work, we propose Adversarial Diffusion Reconstruction Distance (ADRD), a detection framework that models diffusion reconstruction as a dynamic response process rather than a fixed descriptor. ADRD actively probes the reconstruction behavior by introducing perturbation in latent space and measuring how reconstruction deviations respond under identical perturbations. We empirically observe that real images typically exhibit larger and more variable reconstruction responses, while diffusion-generated images tend to show more stable reconstruction behavior, reflecting differences in their alignment with the diffusion model’s implicit data manifold. By characterizing reconstruction sensitivity instead of absolute reconstruction error, ADRD provides a complementary perspective to existing reconstruction-based detectors. Experimental evaluations on multiple benchmarks suggest that reconstruction response under controlled perturbations constitutes a meaningful signal for diffusion-generated image detection. The code is available at https://github.com/ezell-chou/adrd

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Image Processing, Computer-AssistedAlgorithmsGenerative Artificial IntelligenceDiffusion

Autor:innen und Institutionen

  • School of Internet, Jiaxing Vocational and Technical College, Jiaxing, Zhejiang, China.
  • Jiaxing Municipal Public Security Bureau, Jiaxing, Zhejiang, China.
  • Binjiang Institute of Zhejiang University, Hangzhou, Zhejiang, China.

ORCID-Profile: Yi Zhou, Changting Lin, Shuangxi Chen

Förderung

  • Education Department of Zhejiang Province
  • Pioneer and Leading Goose R&D Program of Zhejiang
  • Scientific Research Foundation of Jiaxing Vocational and Technical College
  • Jiaxing Public Welfare Research Project

Zitieren

APA

Zhou, Y., Hu, X., Tong, J., Lin, C., Sun, M., & Chen, S. (2026). ADRD: Detecting diffusion-generated images via adversarial perturbation induced reconstruction discrepancy. PLOS One, 21(7), e0350655. https://doi.org/10.1371/journal.pone.0350655

Vancouver

Zhou Y, Hu X, Tong J, Lin C, Sun M, Chen S. ADRD: Detecting diffusion-generated images via adversarial perturbation induced reconstruction discrepancy. PLOS One. 2026;21(7):e0350655. doi:10.1371/journal.pone.0350655

BibTeX
@article{Zhou2026,
  author = {Zhou, Yi and Hu, Xiangwei and Tong, Jun and Lin, Changting and Sun, Mei and Chen, Shuangxi},
  title = {ADRD: Detecting diffusion-generated images via adversarial perturbation induced reconstruction discrepancy},
  journal = {PLOS One},
  year = {2026},
  volume = {21},
  number = {7},
  pages = {e0350655},
  doi = {10.1371/journal.pone.0350655}
}
Kostenlos starten →