This repository contains code for the paper Copy, Right? A Testing Framework for Copyright Protection of Deep Learning Models (S&P'22).

The code is run successfully using Python 3.6.10 and Tensorflow 2.2.0.

We recommend using conda to install the tensorflow-gpu environment:

$ conda create -n tf2-gpu tensorflow-gpu==2.2.0
$ conda activate tf2-gpu

To run code in the jupyter notebook, you should add the kernel:

$ pip install ipykernel
$ python -m ipykernel install --name tf2-gpu

• DeepJudge: model similarity metrics and test case generation methods.
• train_models: train clean models and suspect models.
• watermarking-whitebox: a TF2 implementation of [1]. (Keras version)
• watermarking-blackbox: a TF2 implementation of [2].
• fingerprinting-blackbox: a TF2 implementation of [3].

Reference:

[1] Uchida et al. "Embedding watermarks into deep neural networks." ICMR 2017.

[2] Zhang et al. "Protecting intellectual property of deep neural networks with watermarking." AisaCCS 2018.

[3] Cao et al. "IPGuard: Protecting intellectual property of deep neural networks via fingerprinting the classification boundary." AsiaCCS 2021.

See the README.md in each directory.