Conference paper · 2025 · Open access

Differentially Private Fine-Tuning of Diffusion Models

Yu‐Lin Tsai, Yizhe Li, Chia-Mu Yu, Xuebin Ren, Po-Yu Chen, Zekai Chen, Francois Buet-Golfouse

IEEE/CVF International Conference on Computer Vision (ICCV), 4561–4571 · Published 19 October 2025 · DOI 10.1109/iccv51701.2025.00434

Open PDFView at the publisherCite this paper

In brief

“Differentially Private Fine-Tuning of Diffusion Models” is a 2025 conference paper by Yu‐Lin Tsai and 6 co-authors, published in IEEE/CVF International Conference on Computer Vision (ICCV). OpenAlex counts 26 citations as of 19 November 2026. It is open access: the full text is free to read at arXiv (Cornell University).

Citations
26
Most cited in
2026 (20)
Field-weighted impact
5.19
References
59
Authors
7

Abstract

The integration of Differential Privacy (DP) with diffusion models (DMs) presents a promising yet challenging frontier, particularly due to the substantial memorization capabilities of DMs that pose significant privacy risks. Differential privacy offers a rigorous framework for safeguarding individual data points during model training, with Differential Privacy Stochastic Gradient Descent (DP-SGD) being a prominent implementation. Diffusion method decomposes image generation into iterative steps, theoretically aligning well with DP's incremental noise addition. Despite the natural fit, the unique architecture of DMs necessitates tailored approaches to effectively balance privacy-utility trade-off. Recent developments in this field have highlighted the potential for generating high-quality synthetic data by pre-training on public data (i.e., ImageNet) and fine-tuning on private data, however, there is a pronounced gap in research on optimizing the trade-offs involved in DP settings, particularly concerning parameter efficiency and model scalability. Our work addresses this by proposing a parameter-efficient fine-tuning strategy optimized for private diffusion models, which minimizes the number of trainable parameters to enhance the privacy-utility trade-off. We empirically demonstrate that our method achieves state-of-the-art performance in DP synthesis, significantly surpassing previous benchmarks on widely studied datasets (e.g., with only 0.47M trainable parameters, achieving a more than 35% improvement over the previous state-of-the-art with a small privacy budget on the CelebA-64 dataset). Anonymous codes available at https://anonymous.4open.science/r/DP-LORA-F02F.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Tsai, Y., Li, Y., Yu, C.-M., Ren, X., Chen, P.-Y., Chen, Z., & Buet-Golfouse, F. (2025). Differentially Private Fine-Tuning of Diffusion Models. IEEE/CVF International Conference on Computer Vision (ICCV), 4561–4571. https://doi.org/10.1109/iccv51701.2025.00434

MLA 9

Tsai, Yu‐Lin, et al. “Differentially Private Fine-Tuning of Diffusion Models.” IEEE/CVF International Conference on Computer Vision (ICCV), 2025, pp. 4561–4571. https://doi.org/10.1109/iccv51701.2025.00434.

Chicago

Tsai, Yu‐Lin, Yizhe Li, Chia-Mu Yu, Xuebin Ren, Po-Yu Chen, Zekai Chen, and Francois Buet-Golfouse. 2025. “Differentially Private Fine-Tuning of Diffusion Models.” IEEE/CVF International Conference on Computer Vision (ICCV): 4561–4571. https://doi.org/10.1109/iccv51701.2025.00434.

Harvard

Tsai, Y. et al. (2025) ‘Differentially Private Fine-Tuning of Diffusion Models’, IEEE/CVF International Conference on Computer Vision (ICCV), pp. 4561–4571. doi:10.1109/iccv51701.2025.00434.

IEEE

Y. Tsai et al., “Differentially Private Fine-Tuning of Diffusion Models,” IEEE/CVF International Conference on Computer Vision (ICCV), pp. 4561–4571, 2025, doi: 10.1109/iccv51701.2025.00434.

BibTeX

@inproceedings{tsai2025differentially,
  title     = {{Differentially Private Fine-Tuning of Diffusion Models}},
  author    = {Tsai, Yu‐Lin and Li, Yizhe and Yu, Chia-Mu and Ren, Xuebin and Chen, Po-Yu and Chen, Zekai and Buet-Golfouse, Francois},
  booktitle = {IEEE/CVF International Conference on Computer Vision (ICCV)},
  year      = {2025},
  pages     = {4561--4571},
  doi       = {10.1109/iccv51701.2025.00434},
  url       = {https://doi.org/10.1109/iccv51701.2025.00434}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Citations by year

YearCitationsShare of the best year
202311
202555
20262020

Its field-weighted citation impact is 5.19: it has been cited 5.19 times as often as the average work of the same field, type and year (1.00 is the world average).

Topic and keywords

Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data

  • differential privacy
  • diffusion models
  • parameter-efficient fine-tuning
  • DP-SGD
  • privacy-utility trade-off
  • memorization
  • LoRa
  • image generation

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Tsai, Y., Li, Y., Yu, C.-M., Ren, X., Chen, P.-Y., Chen, Z., & Buet-Golfouse, F. (2025). Differentially Private Fine-Tuning of Diffusion Models. IEEE/CVF International Conference on Computer Vision (ICCV), 4561–4571. https://doi.org/10.1109/iccv51701.2025.00434

How many times has this paper been cited?

26 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at arXiv (Cornell University): sign in and open the PDF from this page.

Who wrote this paper?

Yu‐Lin Tsai, Yizhe Li, Chia-Mu Yu, Xuebin Ren, Po-Yu Chen, Zekai Chen, Francois Buet-Golfouse.

What is the DOI of this paper?

10.1109/iccv51701.2025.00434, permanently at https://doi.org/10.1109/iccv51701.2025.00434. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.