Unique Security and Privacy Threats of Large Language Models: A Comprehensive Survey
ACM Computing Surveys · 31.03.2026 · 58 (4) · 1-36
DOI: 10.1145/3764113 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Computing Surveys, online erschienen am 06.10.2025. Verfasst von Shang Wang und 6 weiteren Autor:innen. Die Arbeit wurde bisher 46-mal zitiert (Crossref) und nennt 151 Literaturangaben. Gefördert von: Science and Technology Development Fund of Macau SAR, China.
Abstract (Originalsprache)
With the rapid development of artificial intelligence, large language models (LLMs) have made remarkable advancements in natural language processing. These models are trained on vast datasets to exhibit powerful language understanding and generation capabilities across various applications, including chatbots and agents. However, LLMs have revealed a variety of privacy and security issues throughout their life cycle, drawing significant academic and industrial attention. Moreover, the risks faced by LLMs differ significantly from those encountered by traditional language models. Given that current surveys lack a clear taxonomy of unique threat models across diverse scenarios, we emphasize the unique privacy and security threats associated with four specific scenarios: pre-training, fine-tuning, deployment, and LLM-based agents. Addressing the characteristics of each risk, this survey outlines and analyzes potential countermeasures. Research on attack and defense situations can offer feasible research directions, enabling more areas to benefit from LLMs.
Förderung
- Science and Technology Development Fund of Macau SAR, China
Zitieren
APA
Wang, S., Zhu, T., Liu, B., Ding, M., Ye, D., Zhou, W., & Yu, P. (2026). Unique Security and Privacy Threats of Large Language Models: A Comprehensive Survey. ACM Computing Surveys, 58(4), 1-36. https://doi.org/10.1145/3764113
Vancouver
Wang S, Zhu T, Liu B, Ding M, Ye D, Zhou W, et al. Unique Security and Privacy Threats of Large Language Models: A Comprehensive Survey. ACM Computing Surveys. 2026;58(4):1-36. doi:10.1145/3764113
BibTeX
@article{Wang2026,
author = {Wang, Shang and Zhu, Tianqing and Liu, Bo and Ding, Ming and Ye, Dayong and Zhou, Wanlei and Yu, Philip},
title = {Unique Security and Privacy Threats of Large Language Models: A Comprehensive Survey},
journal = {ACM Computing Surveys},
year = {2026},
volume = {58},
number = {4},
pages = {1--36},
doi = {10.1145/3764113}
}
EUVIMED