Deciphering the Chaos: Enhancing Jailbreak Attacks via Adversarial Prompt Translation

TL;DR

This paper introduces a novel method to translate chaotic adversarial prompts into coherent natural language, significantly improving the success rate of jailbreak attacks on various large language models by effectively capturing and transferring semantic vulnerabilities.

Contribution

It is the first to analyze the semantic meaning in garbled adversarial prompts and translate them into human-readable prompts, enhancing transferability and attack success.

Findings

Achieves 81.8% success rate on HarmBench with 10 queries

Over 90% attack success rate on Llama-2-Chat models

Outperforms state-of-the-art methods by large margins

Abstract

Automatic adversarial prompt generation provides remarkable success in jailbreaking safely-aligned large language models (LLMs). Existing gradient-based attacks, while demonstrating outstanding performance in jailbreaking white-box LLMs, often generate garbled adversarial prompts with chaotic appearance. These adversarial prompts are difficult to transfer to other LLMs, hindering their performance in attacking unknown victim models. In this paper, for the first time, we delve into the semantic meaning embedded in garbled adversarial prompts and propose a novel method that "translates" them into coherent and human-readable natural language adversarial prompts. In this way, we can effectively uncover the semantic information that triggers vulnerabilities of the model and unambiguously transfer it to the victim model, without overlooking the adversarial information hidden in the garbled text, to enhance jailbreak attacks. It also offers a new approach to discovering effective designs for jailbreak prompts, advancing the understanding of jailbreak attacks. Experimental results demonstrate that our method significantly improves the success rate of jailbreak attacks against various safety-aligned LLMs and outperforms state-of-the-arts by large margins. With at most 10 queries, our method achieves an average attack success rate of 81.8% in attacking 7 commercial closed-source LLMs, including GPT and Claude-3 series, on HarmBench. Our method also achieves over 90% attack success rates against Llama-2-Chat models on AdvBench, despite their outstanding resistance to jailbreak attacks. Code at: https://github.com/qizhangli/Adversarial-Prompt-Translator.