Jailbreaking? One Step Is Enough!

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zheng, Weixiong, Zeng, Peijian, Li, Yiwei, Wu, Hongyan, Lin, Nankai, Chen, Junhao, Yang, Aimin, Zhou, Yongmei
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915068831072256
author Zheng, Weixiong
Zeng, Peijian
Li, Yiwei
Wu, Hongyan
Lin, Nankai
Chen, Junhao
Yang, Aimin
Zhou, Yongmei
author_facet Zheng, Weixiong
Zeng, Peijian
Li, Yiwei
Wu, Hongyan
Lin, Nankai
Chen, Junhao
Yang, Aimin
Zhou, Yongmei
contents Large language models (LLMs) excel in various tasks but remain vulnerable to jailbreak attacks, where adversaries manipulate prompts to generate harmful outputs. Examining jailbreak prompts helps uncover the shortcomings of LLMs. However, current jailbreak methods and the target model's defenses are engaged in an independent and adversarial process, resulting in the need for frequent attack iterations and redesigning attacks for different models. To address these gaps, we propose a Reverse Embedded Defense Attack (REDA) mechanism that disguises the attack intention as the "defense". intention against harmful content. Specifically, REDA starts from the target response, guiding the model to embed harmful content within its defensive measures, thereby relegating harmful content to a secondary role and making the model believe it is performing a defensive task. The attacking model considers that it is guiding the target model to deal with harmful content, while the target model thinks it is performing a defensive task, creating an illusion of cooperation between the two. Additionally, to enhance the model's confidence and guidance in "defensive" intentions, we adopt in-context learning (ICL) with a small number of attack examples and construct a corresponding dataset of attack examples. Extensive evaluations demonstrate that the REDA method enables cross-model attacks without the need to redesign attack strategies for different models, enables successful jailbreak in one iteration, and outperforms existing methods on both open-source and closed-source models.
format Preprint
id arxiv_https___arxiv_org_abs_2412_12621
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Jailbreaking? One Step Is Enough!
Zheng, Weixiong
Zeng, Peijian
Li, Yiwei
Wu, Hongyan
Lin, Nankai
Chen, Junhao
Yang, Aimin
Zhou, Yongmei
Computation and Language
Large language models (LLMs) excel in various tasks but remain vulnerable to jailbreak attacks, where adversaries manipulate prompts to generate harmful outputs. Examining jailbreak prompts helps uncover the shortcomings of LLMs. However, current jailbreak methods and the target model's defenses are engaged in an independent and adversarial process, resulting in the need for frequent attack iterations and redesigning attacks for different models. To address these gaps, we propose a Reverse Embedded Defense Attack (REDA) mechanism that disguises the attack intention as the "defense". intention against harmful content. Specifically, REDA starts from the target response, guiding the model to embed harmful content within its defensive measures, thereby relegating harmful content to a secondary role and making the model believe it is performing a defensive task. The attacking model considers that it is guiding the target model to deal with harmful content, while the target model thinks it is performing a defensive task, creating an illusion of cooperation between the two. Additionally, to enhance the model's confidence and guidance in "defensive" intentions, we adopt in-context learning (ICL) with a small number of attack examples and construct a corresponding dataset of attack examples. Extensive evaluations demonstrate that the REDA method enables cross-model attacks without the need to redesign attack strategies for different models, enables successful jailbreak in one iteration, and outperforms existing methods on both open-source and closed-source models.
title Jailbreaking? One Step Is Enough!
topic Computation and Language
url https://arxiv.org/abs/2412.12621