🔥SEC-LAB-PPDL Codebase 2026

🔥🔥🔥An open-source Privacy-Preserving Deep Learning framework by SEC LAB.🔥🔥🔥

• Multiple MPC and HE protocols
• One-click CPU / GPU switching
• One-click VOLE / IKNP OT primitive switching

See cpp_coding_style.md for C++ coding conventions.

1. Install Dependencies

SEAL

cd Extern
git clone https://github.com/microsoft/SEAL.git && cd SEAL
git apply ../patch/seal-ckks.patch
cmake -S . -B build -DSEAL_USE_INTEL_HEXL=ON
cmake --build build -j && cd ../..

HEXL

cd Extern
git clone https://github.com/intel/hexl.git && cd hexl
cmake -S . -B build -DCMAKE_INSTALL_PREFIX=hexl_build
cmake --build build -j && cmake --install build && cd ../..

emp-ot & emp-tool

cd Extern && . ./build-ot.sh && cd ..

phantom-fhe (GPU, optional)

Skip this if GPU is not needed (use -DUSE_HE_GPU=OFF when building).

cd Extern
git clone https://github.com/encryptorion-lab/phantom-fhe.git && cd phantom-fhe
git apply ../patch/phantom.patch
cmake -S . -B build -DCMAKE_CUDA_ARCHITECTURES=native -DCMAKE_INSTALL_PREFIX=build_phantom
cmake --build build --target install --parallel && cd ../..

You may need to add the following to ~/.bashrc and run source ~/.bashrc:

export CUDA_HOME=/usr/local/cuda
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64
export PATH=$PATH:$CUDA_HOME/bin

2. Build

cmake -S . -B build          # add -DUSE_HE_GPU=ON for GPU support
cmake --build build -j

3. Run Tests

Tests use a two-party protocol. Start Server (r=1) before Client (r=2) with the same port.

PrivCirNet

test_cir_linear and test_cir_conv are reference implementations for PrivCirNet, testing Block Circulant Linear and Convolution layers respectively. More projects will be added.

# Block Circulant Linear
./build/Test/test_cir_linear r=1 p=1234   # Terminal 1 (Server)
./build/Test/test_cir_linear r=2 p=1234   # Terminal 2 (Client)

# Block Circulant Convolution
./build/Test/test_cir_conv r=1 p=1234     # Terminal 1 (Server)
./build/Test/test_cir_conv r=2 p=1234     # Terminal 2 (Client)

Citation

If you use this codebase, please cite both of the following:

@article{xu2024privcirnet,
  title={Privcirnet: Efficient private inference via block circulant transformation},
  author={Xu, Tianshi and Wu, Lemeng and Wang, Runsheng and Li, Meng},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={111802--111831},
  year={2024}
}

@inproceedings{xu2025breaking,
  title={Breaking the layer barrier: Remodeling private transformer inference with hybrid {CKKS} and {MPC}},
  author={Xu, Tianshi and Lu, Wen-jie and Yu, Jiangrui and Chen, Yi and Lin, Chenqi and Wang, Runsheng and Li, Meng},
  booktitle={34th USENIX Security Symposium (USENIX Security 25)},
  pages={2653--2672},
  year={2025}
}