Update README.md

README.md

@@ -1 +1,37 @@
-# PrivacyJailbreak
+# <img src="/Users/xiaohong/Documents/Github/PrivacyJailbreak/img/logo.png" style="zoom:50%;" />PIG: Privacy Jailbreak Attack on LLMs via Gradient-based Iterative In-Context Optimization
+
+> This repository contains the official code implementation of our paper: [![arXiv: paper](https://img.shields.io/badge/arXiv-paper-red.svg)](https://arxiv.org/abs/xxx)
+
+![PIG](/Users/xiaohong/Documents/Github/PrivacyJailbreak/img/PIG.png)
+
+## Setup
+
+First, create a virtual environment using Anaconda:
+
+```python
+conda create -n pig python=3.9.19
+conda activate pig
+```
+
+Second, you need to install the necessary dependencies:
+
+```python
+pip install -r requirements.txt
+```
+
+## Datasets
+
+You can download the Enron Email dataset and TrustLLM dataset [here](https://drive.google.com/drive/folders/16Th72F_QcxRAryOIk9L2t0oIps1xnHGW) and place them under the `./data` directory.
+
+## Usage
+
+You can run a privacy jailbreak attack using the following steps:
+
+1. First, modify parameters such as `dataset`, `target_model_name`, `attack_model_name`, or `eval_model_name` in script `run.sh`.
+2. Then, execute the privacy jailbreak attack by running `bash run.sh`. Use the `tail` command to monitor the `log` file in real time.
+3. Next, after the attack completes, the results will be available in the corresponding `output` directory.
+4. Finally, evaluate the results using `python eval.py` to compute various metrics such as the Attack Success Rate (ASR).
+
+## Acknowledgements
+
+Our PIG framework is based on [EasyJailbreak](https://github.com/EasyJailbreak/EasyJailbreak). We thank the team for their open-source implementation.