CAVGAN: Unifying Jailbreak and Defense of LLMs via Generative Adversarial Attacks on their Internal Representations Article Swipe

MAKE IMPACT

Xiaohu Li , Yunfeng Ning , Zepeng Bao , Mayi Xu , Jianhao Chen , Tieyun Qian · YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2507.06043

Security alignment enables the Large Language Model (LLM) to gain the protection against malicious queries, but various jailbreak attack methods reveal the vulnerability of this security mechanism. Previous studies have isolated LLM jailbreak attacks and defenses. We analyze the security protection mechanism of the LLM, and propose a framework that combines attack and defense. Our method is based on the linearly separable property of LLM intermediate layer embedding, as well as the essence of jailbreak attack, which aims to embed harmful problems and transfer them to the safe area. We utilize generative adversarial network (GAN) to learn the security judgment boundary inside the LLM to achieve efficient jailbreak attack and defense. The experimental results indicate that our method achieves an average jailbreak success rate of 88.85\% across three popular LLMs, while the defense success rate on the state-of-the-art jailbreak dataset reaches an average of 84.17\%. This not only validates the effectiveness of our approach but also sheds light on the internal security mechanisms of LLMs, offering new insights for enhancing model security The code and data are available at https://github.com/NLPGM/CAVGAN.

Type

preprint

Language

en

Landing Page

http://arxiv.org/abs/2507.06043

PDF

https://arxiv.org/pdf/2507.06043

OA Status

green

OpenAlex ID

https://openalex.org/W4416062077

Raw OpenAlex JSON

OpenAlex ID

https://openalex.org/W4416062077

Canonical identifier for this work in OpenAlex

DOI

https://doi.org/10.48550/arxiv.2507.06043

Digital Object Identifier

Title

CAVGAN: Unifying Jailbreak and Defense of LLMs via Generative Adversarial Attacks on their Internal Representations

Work title

Type

preprint

OpenAlex work type

Language

en

Primary language

Publication year

2025

Year of publication

Publication date

2025-07-08

Full publication date if available

Authors

Xiaohu Li, Yunfeng Ning, Zepeng Bao, Mayi Xu, Jianhao Chen, Tieyun Qian

List of authors in order

Landing page

https://arxiv.org/abs/2507.06043

Publisher landing page

PDF URL

https://arxiv.org/pdf/2507.06043

Direct link to full text PDF

Open access

Yes

Whether a free full text is available

OA status

green

Open access status per OpenAlex

OA URL

https://arxiv.org/pdf/2507.06043

Direct OA link when available

Cited by

0

Total citation count in OpenAlex

Full payload