Phantom: A CUDA-Accelerated Fully Homomorphic Encryption Library

Prerequisites

CUDA >= 11.0
CMake >= 3.20
GCC >= 11.0

CMake Options

CMAKE_CUDA_ARCHITECTURES: Set the CUDA architectures to compile for. For example, A100 uses 80, V100 uses 70, and P100 uses 60.

Features

Native GPU acceleration (for NVIDIA GPUs)
Support word-wise schemes including BGV, BFV, and CKKS (without bootstrapping)
SOTA performance in most operations
Easy to integrate with applications (PPML, etc.)

Usage

Use git to clone this repository recursively (including submodules)
Use CMake to configure and build this library
Look into build/bin and execute binaries
(Optional) Use python bindings (See python/ directory for details)

License

This project (Phantom) is released under GPLv3 license. See COPYING for more information.

Some files contain the modified code from Microsoft SEAL. These codes are released under MIT License. See MIT License for more information.

Citation

If you use Phantom in your research, please cite the following paper:

Early access TDSC version (IEEE Xplore):

@article{10428046,
         author={Yang, Hao and Shen, Shiyu and Dai, Wangchen and Zhou, Lu and Liu, Zhe and Zhao, Yunlei},
         journal={IEEE Transactions on Dependable and Secure Computing}, 
         title={Phantom: A CUDA-Accelerated Word-Wise Homomorphic Encryption Library}, 
         year={2024},
         volume={},
         number={},
         pages={1-12},
         doi={10.1109/TDSC.2024.3363900}
}

IACR ePrint version (Cryptology ePrint Archive):

@misc{cryptoeprint:2023/049,
      author = {Hao Yang and Shiyu Shen and Wangchen Dai and Lu Zhou and Zhe Liu and Yunlei Zhao},
      title = {Phantom: A CUDA-Accelerated Word-Wise Homomorphic Encryption Library},
      howpublished = {Cryptology ePrint Archive, Paper 2023/049},
      year = {2023},
      doi = {10.1109/TDSC.2024.3363900},
      note = {\url{https://eprint.iacr.org/2023/049}},
      url = {https://eprint.iacr.org/2023/049}
}

If you are exploring BFV optimizations, please also cite the following paper:

@misc{cryptoeprint:2023/1429,
      author = {Shiyu Shen and Hao Yang and Wangchen Dai and Lu Zhou and Zhe Liu and Yunlei Zhao},
      title = {Leveraging GPU in Homomorphic Encryption: Framework Design and Analysis of BFV Variants},
      howpublished = {Cryptology ePrint Archive, Paper 2023/1429},
      year = {2023},
      note = {\url{https://eprint.iacr.org/2023/1429}},
      url = {https://eprint.iacr.org/2023/1429}
}

Roadmap

We are planning to support the following features in the future:

[] support bit-wise schemes TFHE/FHEW
[] support scheme switching between word-wise schemes and bit-wise schemes
[] support bootstrapping for BFV/BGV/CKKS

petitioner / cude-fhe