The official implementation of our pre-print paper "AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models".
[Paper]

The aligned Large Language Models (LLMs) are powerful language understanding and decision-making tools that are created through extensive alignment with human feedback. However, these large models remain susceptible to jailbreak attacks, where adversaries manipulate prompts to elicit malicious outputs that should not be given by aligned LLMs. Investigating jailbreak prompts can lead us to delve into the limitations of LLMs and further guide us to secure them. Unfortunately, existing jailbreak techniques suffer from either (1) scalability issues, where attacks heavily rely on manual crafting of prompts, or (2) stealthiness problems, as attacks depend on token-based algorithms to generate prompts that are often semantically meaningless, making them susceptible to detection through basic perplexity testing. In light of these challenges, we intend to answer this question: Can we develop an approach that can automatically generate stealthy jailbreak prompts? In this paper, we introduce AutoDAN, a novel jailbreak attack against aligned LLMs. AutoDAN can automatically generate stealthy jailbreak prompts by the carefully designed hierarchical genetic algorithm. Extensive evaluations demonstrate that AutoDAN not only automates the process while preserving semantic meaningfulness, but also demonstrates superior attack strength in cross-model transferability, and cross-sample universality compared with the baseline. Moreover, we also compare AutoDAN with perplexity-based defense methods and show that AutoDAN can bypass them effectively.

2023/10/11 We have released the quick implementation of AutoDAN. More instructions will be updated soon.
2023/10/03 We have released our paper.

• Get code

git clone https://github.com/SheltonLiu-N/AutoDAN.git

• Build environment

cd AutoDAN
conda create -n AutoDAN python=3.9
conda activate AutoDAN
pip install -r requirements.txt

• Download LLMs (Should have permissions when downloading Llama2, see this page)

cd start
python download_models.py

• AutoDAN

python AutoDAN.py

• With GPT mutation

python AutoDAN.py --API_key <your openai API key>

• Runing in low GPU memory (≈24GB) (Will cost more time)

python AutoDAN.py --API_key <your openai API key> --low_memory 1

• Result (Snapshot from Chat_with_Llama2, Llama2-7b)

Some of our codes are built upon llm-attack.

@misc{liu2023autodan,
      title={AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models}, 
      author={Xiaogeng Liu and Nan Xu and Muhao Chen and Chaowei Xiao},
      year={2023},
      eprint={2310.04451},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}