Stealix: Model Stealing via Prompt Evolution

🔥 Official implementation of "Stealix: Model Stealing via Prompt Evolution" (ICML 2025)

---

Table of Contents

• Overview
• Getting Started
• Usage

---

Overview

This repository contains the implementation of "Stealix: Model Stealing via Prompt Evolution," the first victim-aware prompt optimization method for model stealing.

---

Getting Started

Follow the steps below to set up the required Python environment and dependencies:

1. Create a Conda Environment from environment.yml

   conda env create -f environment.yml

2. Activate the Environment

   conda activate stealix

---

Usage

The project includes four main scripts in the script/ folder to run the complete pipeline:

1. Prepare Seed Images

   bash script/0_prepare_seeds.sh

2. Train Victim Model

   bash script/1_train_victim_model.sh

3. Prompt Evolution

   bash script/2_prompt_evolution.sh

4. Train Stealix Model

   bash script/3_train_stealix_model.sh

Citation

If you find our work useful, please star this repo and cite:

@inproceedings{zhuang2025stealix,
        title={Stealix: Model Stealing via Prompt Evolution},
        author={Zhuang, Zhixiong and Wang, Hui-Po and Nicolae, Maria-Irina and Fritz, Mario},
        booktitle={International Conference on Machine Learning (ICML)},
        year={2025}
}

License

This project is open-sourced under the AGPL-3.0 license. See the LICENSE file for details.

For a list of other open source components included in this project, see the file 3rd-party-licenses.txt.

Purpose of the project

This software is a research prototype, solely developed for and published as part of the publication cited above.

Contact

Please feel free to open an issue or contact personally if you have questions, need help, or need explanations. Don't hesitate to write an email to the following email address: [email protected]