Official repository for HETAL: Efficient Privacy-preserving Transfer Learning with Homomorphic Encryption (ICML'23 Oral) by Seewoo Lee1*, Garam Lee2, Jung Woo Kim2, Junbum Shin2, and Mun-Kyu Lee3**.
1University of California, Berkeley 2CryptoLab 3Inha University
* Work done at CryptoLab
** Corresponding author
You can read the paper here: PMLR Link
For the updates after publication, see arXiv version: arXiv Link
This repository uses CPU version of HEaaN library. To use GPU acceleration, please contact stat@cryptolab.co.kr.
- OS: Linux
- Python: 3.8
- Recommended Memory: 32GB (Minimum : 16GB)
First install pipenv. Then run the following shell commands.
All the required packages including heaan
and heaan_sdk
will be installed using existing whl
files (heaan_sdk-0.2.0-cp38-cp38-linux_x86_64.whl
and heaan-0.1.0+cpu-cp38-cp38-linux_x86_64.whl
).
pipenv --python 3.8
pipenv shell
pipenv install
See README.md
files in src/hetal
and src/benchmark
directories for guides.
Use the following Bibtex entry for citation.
@article{lee2023hetal,
title={HETAL: Efficient Privacy-preserving Transfer Learning with Homomorphic Encryption},
author={Lee, Seewoo and Lee, Garam and Kim, Jung Woo and Shin, Junbum and Lee, Mun-Kyu},
journal={ICML},
year={2023}
}
This is available for the non-commercial purpose only. See LICENSE
for details.
HETAL is now integrated into CryptoLab's new product, AutoFHE. It is available in autofhe.com.