StreamGuard

Getting Start

Get Code

git clone https://github.com/Lilaizhen/StreamGuard.git

Build Environment

cd StreamGuard
conda create -n StreamGuard python=3.10
conda activate StreamGuard
pip install -r requirements.txt

To start,

cd exp
python defense.py --model_name [YOUR_MODEL_NAME] --attacker [YOUR_ATTACKER_NAME] --defender [YOUR_DEFENDER_NAME] --GPT_API [YOUR_OPENAI_API]

Current Supports:

• Model Name: llama3.1, qwen2.5.

• Attacker: GCG, AutoDAN, DeepInception, AdvBench and your customized dataset.

• Defender: StreamGuard, PPL, Self-Exam, Paraphrase, Retokenization, Self-Reminder, ICD.

Don't forget to add your openai api to get harmful scores. If you only want to get ASR, you can

python defense.py --model_name [YOUR_MODEL_NAME] --attacker [YOUR_ATTACKER_NAME] --defender [YOUR_DEFENDER_NAME] --disable_GPT_judge

cd exp
python defense.py --model_name [YOUR_MODEL_NAME] --attacker Just-Eval --defender [YOUR_DEFENDER_NAME] --GPT_API [YOUR_OPENAI_API]

cd exp
python finetune.py --model_name [YOUR_MODEL_NAME] --GPT_API [YOUR_OPENAI_API]

Acknowledgements

Huge thanks to the following repos that greatly help our implementation:

• https://github.com/uw-nsl/SafeDecoding
• https://github.com/SheltonLiu-N/AutoDAN