Preprint · 2025 · Open access

SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks

Kaiyuan Zhang, Siyuan Cheng, Hanxi Guo, Yuetian Chen, Zian Su, Shengwei An, Yuntao Du, Fleming, Charles, Ashish Kundu, Xiangyu Zhang, Ninghui Li

arXiv (Cornell University) · Published 12 June 2025 · DOI 10.48550/arxiv.2506.10424

Open PDFView at the publisherCite this paper

In brief

“SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks” is a 2025 preprint by Kaiyuan Zhang and 10 co-authors, published in arXiv (Cornell University). OpenAlex counts 22 citations as of 19 November 2026. It is open access under a CC BY licence: the full text is free to read at arXiv (Cornell University).

Citations
22
Most cited in
2026 (18)
References
76
Authors
11

Abstract

Large language models (LLMs) have achieved remarkable success and are widely adopted for diverse applications. However, fine-tuning these models often involves private or sensitive information, raising critical privacy concerns. In this work, we conduct the first comprehensive study evaluating the vulnerability of fine-tuned LLMs to membership inference attacks (MIAs). Our empirical analysis demonstrates that MIAs exploit the loss reduction during fine-tuning, making them highly effective in revealing membership information. These findings motivate the development of our defense. We propose SOFT (\textbf{S}elective data \textbf{O}bfuscation in LLM \textbf{F}ine-\textbf{T}uning), a novel defense technique that mitigates privacy leakage by leveraging influential data selection with an adjustable parameter to balance utility preservation and privacy protection. Our extensive experiments span six diverse domains and multiple LLM architectures and scales. Results show that SOFT effectively reduces privacy risks while maintaining competitive model performance, offering a practical and scalable solution to safeguard sensitive information in fine-tuned LLMs.

Cite this paper

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

APA 7

Zhang, K., Cheng, S., Guo, H., Chen, Y., Su, Z., An, S., Du, Y., Charles, F., Kundu, A., Zhang, X., & Li, N. (2025). SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2506.10424

MLA 9

Zhang, Kaiyuan, et al. “SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks.” arXiv (Cornell University), 2025. https://doi.org/10.48550/arxiv.2506.10424.

Chicago

Zhang, Kaiyuan, Siyuan Cheng, Hanxi Guo, Yuetian Chen, Zian Su, Shengwei An, Yuntao Du, et al. 2025. “SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks.” arXiv (Cornell University). https://doi.org/10.48550/arxiv.2506.10424.

Harvard

Zhang, K. et al. (2025) ‘SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks’, arXiv (Cornell University). doi:10.48550/arxiv.2506.10424.

IEEE

K. Zhang et al., “SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks,” arXiv (Cornell University), 2025, doi: 10.48550/arxiv.2506.10424.

BibTeX

@misc{zhang2025soft,
  title     = {{SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks}},
  author    = {Zhang, Kaiyuan and Cheng, Siyuan and Guo, Hanxi and Chen, Yuetian and Su, Zian and An, Shengwei and Du, Yuntao and Charles, Fleming, and Kundu, Ashish and Zhang, Xiangyu and Li, Ninghui},
  howpublished = {arXiv (Cornell University)},
  year      = {2025},
  doi       = {10.48550/arxiv.2506.10424},
  url       = {https://doi.org/10.48550/arxiv.2506.10424}
}

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

Topic and keywords

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

  • membership inference attacks
  • LLM fine-tuning
  • data obfuscation
  • privacy leakage
  • privacy-preserving machine learning
  • privacy-utility trade-off
  • loss reduction
  • SOFT

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?

Zhang, K., Cheng, S., Guo, H., Chen, Y., Su, Z., An, S., Du, Y., Charles, F., Kundu, A., Zhang, X., & Li, N. (2025). SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2506.10424

How many times has this paper been cited?

22 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?

Kaiyuan Zhang, Siyuan Cheng, Hanxi Guo, Yuetian Chen, Zian Su, Shengwei An, Yuntao Du, Fleming, Charles, Ashish Kundu, Xiangyu Zhang, Ninghui Li.

What is the DOI of this paper?

10.48550/arxiv.2506.10424, permanently at https://doi.org/10.48550/arxiv.2506.10424. 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.