Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey
ACM Computing Surveys · 31.10.2026 · 58 (13) · 1-39
DOI: 10.1145/3817114 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Computing Surveys, online erschienen am 23.06.2026. Verfasst von Tiansheng Huang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 295 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NSF CISE, CISCO Edge AI program.
Abstract (Originalsprache)
Recent research demonstrates that the nascent fine-tuning-as-a-service business model exposes serious safety concerns: fine-tuning with a few harmful data or even seemingly benign data uploaded from the users can compromise the safety alignment of the model. The attack, known as harmful fine-tuning attack, has generated broad research interests in both academia and industry. In this article, we first systematically formulate the threat model and basic assumptions of harmful fine-tuning. Then, we provide a comprehensive review of harmful fine-tuning from three fundamental perspectives: attack setting, defense design, and evaluation methodology. First, we present the threat model of the problem and introduce the harmful fine-tuning attack. Next, we systematically survey representative attacks, defense methods, and mechanical interpretability in the existing literature. Finally, we introduce the evaluation methodology and outline future research directions, which can serve as guidelines and crucial perspectives for the future development of the subject. We also maintain a curated list of relevant papers, which are made accessible at https://github.com/git-disl/awesome_LLM-harmful-fine-tuning-papers . Disclaimer: This document contains content that some may find disturbing or offensive, including content that is hateful or violent.
Förderung
- NSF CISE
- CISCO Edge AI program
Zitieren
APA
Huang, T., Hu, S., Ilhan, F., Tekin, S. F., & Liu, L. (2026). Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey. ACM Computing Surveys, 58(13), 1-39. https://doi.org/10.1145/3817114
Vancouver
Huang T, Hu S, Ilhan F, Tekin SF, Liu L. Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey. ACM Computing Surveys. 2026;58(13):1-39. doi:10.1145/3817114
BibTeX
@article{Huang2026,
author = {Huang, Tiansheng and Hu, Sihao and Ilhan, Fatih and Tekin, Selim Furkan and Liu, Ling},
title = {Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey},
journal = {ACM Computing Surveys},
year = {2026},
volume = {58},
number = {13},
pages = {1--39},
doi = {10.1145/3817114}
}
EUVIMED