.png){kind=icon}
返回
Dialogue Injection Attack: Jailbreaking LLMs Through Context Manipulation
.png){kind=icon}
PRE
AI
DOI：10.1109/TIFS.2026.3657898.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Large language models (LLMs) have demonstrated significant utility in a wide range of applications; however, their deployment is plagued by security vulnerabilities, notably jailbreak attacks. These attacks manipulate LLMs to generate harmful or unethical content by crafting adversarial prompts. While much of the current research on jailbreak attacks has focused on single-turn interactions, it has largely overlooked the impact of historical dialogues on model behavior. Although recent studies have explored multi-turn jailbreak attacks, they generally assume that the attacker can only manipulate the user prompt. In contrast, we highlight that an attacker can also control the model’s previous outputs. To this end, we introduce $\textsf {DIA}$ , a new paradigm that leverages fabricated dialogue history to enhance jailbreak effectiveness. $\textsf {DIA}$ operates in a black-box setting, requiring only access to the chat API or knowledge of the LLM’s chat template. We propose two methods for constructing adversarial historical dialogues: one adapts gray-box prefilling attacks, and the other exploits deferred responses. Our experiments demonstrate that $\textsf {DIA}$ achieves state-of-the-art attack success rates on recent LLMs, including Llama-3.1 and GPT-4o. Additionally, we show that DIA can bypass 6 different defense mechanisms, highlighting its robustness. Code is available at https://github.com/meng-wenlong/DIA
Keyword:
Jailbreak attacks
dialogue injection
safety alignment
期刊
IF:
8
论文数:
5.3K
被引数:
2.3W
机构
引用论文
P. Chao, A. Robey, E. Dobriban, H. Hassani, G. J. Pappas, E. Wong, Jailbreaking black box large language models in twenty queries, in: 2025 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML), IEEE, 2025, pp. 23–42.赵 P., Robey A., Dobriban E., Hassani H., Pappas G. J., Wong E., 在二十次查询中破解黑箱大型语言模型，载于：2025年IEEE安全与可信机器学习会议（SaTML），IEEE，2025年，第23-42页。.png)
IF0
PREAI
分享收藏Efficient Memory Management for Large Language Model Serving with PagedAttention使用PagedAttention服务的大型语言模型的有效内存管理
OAAI
OAAI分享收藏分享收藏A Wolf in Sheep’s Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily披着羊皮的狼：通用嵌套越狱提示可以轻松欺骗大型语言模型OAAI
OAAI分享收藏
查看更多内容