Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2401.06561 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866929631250087936 |
|---|---|
| author | Zhang, Yuqi Ding, Liang Zhang, Lefei Tao, Dacheng |
| author_facet | Zhang, Yuqi Ding, Liang Zhang, Lefei Tao, Dacheng |
| contents | Aligning large language models (LLMs) with human values, particularly when facing complex and stealthy jailbreak attacks, presents a formidable challenge. Unfortunately, existing methods often overlook this intrinsic nature of jailbreaks, which limits their effectiveness in such complex scenarios. In this study, we present a simple yet highly effective defense strategy, i.e., Intention Analysis ($\mathbb{IA}$). $\mathbb{IA}$ works by triggering LLMs' inherent self-correct and improve ability through a two-stage process: 1) analyzing the essential intention of the user input, and 2) providing final policy-aligned responses based on the first round conversation. Notably, $\mathbb{IA}$ is an inference-only method, thus could enhance LLM safety without compromising their helpfulness. Extensive experiments on varying jailbreak benchmarks across a wide range of LLMs show that $\mathbb{IA}$ could consistently and significantly reduce the harmfulness in responses (averagely -48.2% attack success rate). Encouragingly, with our $\mathbb{IA}$, Vicuna-7B even outperforms GPT-3.5 regarding attack success rate. We empirically demonstrate that, to some extent, $\mathbb{IA}$ is robust to errors in generated intentions. Further analyses reveal the underlying principle of $\mathbb{IA}$: suppressing LLM's tendency to follow jailbreak prompts, thereby enhancing safety. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_06561 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Intention Analysis Makes LLMs A Good Jailbreak Defender Zhang, Yuqi Ding, Liang Zhang, Lefei Tao, Dacheng Computation and Language Aligning large language models (LLMs) with human values, particularly when facing complex and stealthy jailbreak attacks, presents a formidable challenge. Unfortunately, existing methods often overlook this intrinsic nature of jailbreaks, which limits their effectiveness in such complex scenarios. In this study, we present a simple yet highly effective defense strategy, i.e., Intention Analysis ($\mathbb{IA}$). $\mathbb{IA}$ works by triggering LLMs' inherent self-correct and improve ability through a two-stage process: 1) analyzing the essential intention of the user input, and 2) providing final policy-aligned responses based on the first round conversation. Notably, $\mathbb{IA}$ is an inference-only method, thus could enhance LLM safety without compromising their helpfulness. Extensive experiments on varying jailbreak benchmarks across a wide range of LLMs show that $\mathbb{IA}$ could consistently and significantly reduce the harmfulness in responses (averagely -48.2% attack success rate). Encouragingly, with our $\mathbb{IA}$, Vicuna-7B even outperforms GPT-3.5 regarding attack success rate. We empirically demonstrate that, to some extent, $\mathbb{IA}$ is robust to errors in generated intentions. Further analyses reveal the underlying principle of $\mathbb{IA}$: suppressing LLM's tendency to follow jailbreak prompts, thereby enhancing safety. |
| title | Intention Analysis Makes LLMs A Good Jailbreak Defender |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2401.06561 |