Bypassing the Safety Training of Open-Source LLMs with Priming Attacks
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2023
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914799487549440 |
|---|---|
| author | Vega, Jason Chaudhary, Isha Xu, Changming Singh, Gagandeep |
| author_facet | Vega, Jason Chaudhary, Isha Xu, Changming Singh, Gagandeep |
| contents | With the recent surge in popularity of LLMs has come an ever-increasing need for LLM safety training. In this paper, we investigate the fragility of SOTA open-source LLMs under simple, optimization-free attacks we refer to as $\textit{priming attacks}$, which are easy to execute and effectively bypass alignment from safety training. Our proposed attack improves the Attack Success Rate on Harmful Behaviors, as measured by Llama Guard, by up to $3.3\times$ compared to baselines. Source code and data are available at https://github.com/uiuc-focal-lab/llm-priming-attacks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2312_12321 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Bypassing the Safety Training of Open-Source LLMs with Priming Attacks Vega, Jason Chaudhary, Isha Xu, Changming Singh, Gagandeep Cryptography and Security Artificial Intelligence Computation and Language Machine Learning With the recent surge in popularity of LLMs has come an ever-increasing need for LLM safety training. In this paper, we investigate the fragility of SOTA open-source LLMs under simple, optimization-free attacks we refer to as $\textit{priming attacks}$, which are easy to execute and effectively bypass alignment from safety training. Our proposed attack improves the Attack Success Rate on Harmful Behaviors, as measured by Llama Guard, by up to $3.3\times$ compared to baselines. Source code and data are available at https://github.com/uiuc-focal-lab/llm-priming-attacks. |
| title | Bypassing the Safety Training of Open-Source LLMs with Priming Attacks |
| topic | Cryptography and Security Artificial Intelligence Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2312.12321 |