Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2410.23678 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910048260718592 |
|---|---|
| author | Chen, Zheng Yao, Buhui |
| author_facet | Chen, Zheng Yao, Buhui |
| contents | Goal hijacking is a type of adversarial attack on Large Language Models (LLMs) where the objective is to manipulate the model into producing a specific, predetermined output, regardless of the user's original input. In goal hijacking, an attacker typically appends a carefully crafted malicious suffix to the user's prompt, which coerces the model into ignoring the user's original input and generating the target response. In this paper, we introduce a novel goal hijacking attack method called Pseudo-Conversation Injection, which leverages the weaknesses of LLMs in role identification within conversation contexts. Specifically, we construct the suffix by fabricating responses from the LLM to the user's initial prompt, followed by a prompt for a malicious new task. This leads the model to perceive the initial prompt and fabricated response as a completed conversation, thereby executing the new, falsified prompt. Following this approach, we propose three Pseudo-Conversation construction strategies: Targeted Pseudo-Conversation, Universal Pseudo-Conversation, and Robust Pseudo-Conversation. These strategies are designed to achieve effective goal hijacking across various scenarios. Our experiments, conducted on two mainstream LLM platforms including ChatGPT and Qwen, demonstrate that our proposed method significantly outperforms existing approaches in terms of attack effectiveness. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_23678 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Goal Hijacking Attack on Large Language Models via Pseudo-Conversation Injection Chen, Zheng Yao, Buhui Computation and Language Goal hijacking is a type of adversarial attack on Large Language Models (LLMs) where the objective is to manipulate the model into producing a specific, predetermined output, regardless of the user's original input. In goal hijacking, an attacker typically appends a carefully crafted malicious suffix to the user's prompt, which coerces the model into ignoring the user's original input and generating the target response. In this paper, we introduce a novel goal hijacking attack method called Pseudo-Conversation Injection, which leverages the weaknesses of LLMs in role identification within conversation contexts. Specifically, we construct the suffix by fabricating responses from the LLM to the user's initial prompt, followed by a prompt for a malicious new task. This leads the model to perceive the initial prompt and fabricated response as a completed conversation, thereby executing the new, falsified prompt. Following this approach, we propose three Pseudo-Conversation construction strategies: Targeted Pseudo-Conversation, Universal Pseudo-Conversation, and Robust Pseudo-Conversation. These strategies are designed to achieve effective goal hijacking across various scenarios. Our experiments, conducted on two mainstream LLM platforms including ChatGPT and Qwen, demonstrate that our proposed method significantly outperforms existing approaches in terms of attack effectiveness. |
| title | Goal Hijacking Attack on Large Language Models via Pseudo-Conversation Injection |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2410.23678 |