AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912339577536512 |
|---|---|
| author | Liu, Xiaogeng Li, Peiran Suh, Edward Vorobeychik, Yevgeniy Mao, Zhuoqing Jha, Somesh McDaniel, Patrick Sun, Huan Li, Bo Xiao, Chaowei |
| author_facet | Liu, Xiaogeng Li, Peiran Suh, Edward Vorobeychik, Yevgeniy Mao, Zhuoqing Jha, Somesh McDaniel, Patrick Sun, Huan Li, Bo Xiao, Chaowei |
| contents | In this paper, we propose AutoDAN-Turbo, a black-box jailbreak method that can automatically discover as many jailbreak strategies as possible from scratch, without any human intervention or predefined scopes (e.g., specified candidate strategies), and use them for red-teaming. As a result, AutoDAN-Turbo can significantly outperform baseline methods, achieving a 74.3% higher average attack success rate on public benchmarks. Notably, AutoDAN-Turbo achieves an 88.5 attack success rate on GPT-4-1106-turbo. In addition, AutoDAN-Turbo is a unified framework that can incorporate existing human-designed jailbreak strategies in a plug-and-play manner. By integrating human-designed strategies, AutoDAN-Turbo can even achieve a higher attack success rate of 93.4 on GPT-4-1106-turbo. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_05295 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs Liu, Xiaogeng Li, Peiran Suh, Edward Vorobeychik, Yevgeniy Mao, Zhuoqing Jha, Somesh McDaniel, Patrick Sun, Huan Li, Bo Xiao, Chaowei Cryptography and Security Artificial Intelligence Machine Learning In this paper, we propose AutoDAN-Turbo, a black-box jailbreak method that can automatically discover as many jailbreak strategies as possible from scratch, without any human intervention or predefined scopes (e.g., specified candidate strategies), and use them for red-teaming. As a result, AutoDAN-Turbo can significantly outperform baseline methods, achieving a 74.3% higher average attack success rate on public benchmarks. Notably, AutoDAN-Turbo achieves an 88.5 attack success rate on GPT-4-1106-turbo. In addition, AutoDAN-Turbo is a unified framework that can incorporate existing human-designed jailbreak strategies in a plug-and-play manner. By integrating human-designed strategies, AutoDAN-Turbo can even achieve a higher attack success rate of 93.4 on GPT-4-1106-turbo. |
| title | AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs |
| topic | Cryptography and Security Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2410.05295 |