Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction

Tong Liu, Yingjie Zhang, Zhe Zhao, Yinpeng Dong, Guozhu Meng +1 more
2/28/2024
cs.CRcs.AI

No official code found. Best implementation: LLM-DRA/DRA (116 stars, MIT, last commit 2024-10-11). 1 more below.

Abstract

In recent years, large language models (LLMs) have demonstrated notable success across various tasks, but the trustworthiness of LLMs is still an open problem. One specific threat is the potential to generate toxic or harmful responses. Attackers can craft adversarial prompts that induce harmful responses from LLMs. In this work, we pioneer a theoretical foundation in LLMs security by identifying bias vulnerabilities within the safety fine-tuning and design a black-box jailbreak method named DRA (Disguise and Reconstruction Attack), which conceals harmful instructions through disguise and prompts the model to reconstruct the original harmful instruction within its completion. We evaluate DRA across various open-source and closed-source models, showcasing state-of-the-art jailbreak success rates and attack efficiency. Notably, DRA boasts a 91.1% attack success rate on OpenAI GPT-4 chatbot.

View on arXivView PDF

Code Implementations(2)

LLM-DRA/DRACommunity100%
MIT

[USENIX Security'24] Official repository of "Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction"

11614PythonJun 4, 20242 years agoMIT
LLM-DRA/llm-dra.github.ioCommunity57%

Demo site of the paper Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction

30CSS, HTMLFeb 26, 20242 years ago

Similar Papers

Fine-Tuning, Quantization, and LLMs: Navigating Unintended Outcomes

94% match2024

PUZZLED: Jailbreaking LLMs through Word-Based Puzzles

92% match1 repo2025

Behind the Mask: Benchmarking Camouflaged Jailbreaks in Large Language Models

91% match2025

Diversity Helps Jailbreak Large Language Models

91% match2024

NeuroBreak: Unveil Internal Jailbreak Mechanisms in Large Language Models

91% match2025

All in How You Ask for It: Simple Black-Box Method for Jailbreak Attacks

91% match1 repo172024

Cite this paper

@article{liu2024making,
  title  = {Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction},
  author = {Tong Liu and Yingjie Zhang and Zhe Zhao and Yinpeng Dong and Guozhu Meng and Kai Chen},
  year   = {2024},
  eprint = {2402.18104},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2402.18104v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2402.18104.svg)](https://codeofpaper.com/paper/2402.18104)

Discussion