Concrete is a TFHE-based compiler that simplifies the development of Fully Homomorphic Encryption (FHE) applications. It targets developers who want to perform computations on encrypted data without deep cryptographic expertise, offering a Python API and GPU acceleration for privacy-preserving services and data breach mitigation.

How It Works

Concrete leverages LLVM to compile Python programs into FHE circuits. It abstracts the complexities of TFHE, allowing users to define functions in Python, specify encrypted inputs, and compile them into executable circuits. This approach enables computations on encrypted data, with automatic parameter selection and performance analysis tools integrated for ease of use.

Quick Start & Requirements

• Installation: pip install concrete-python (CPU) or pip install concrete-python --index-url https://pypi.zama.ai/gpu (GPU). Docker images are also available.
• Prerequisites: Python 3.9-3.12, 8GB RAM (16GB recommended). GPU version requires CUDA ≥ 11.8.
• Resources: Detailed installation and usage examples are available in the documentation.

Highlighted Details

• Python API for FHE operations.
• GPU acceleration support.
• Automatic parameter selection and performance analysis.
• Built on TFHE technology and LLVM.

Maintenance & Community

• Community support is available via Discord.
• Zama offers bounty and grant programs for contributions.

Licensing & Compatibility

• BSD-3-Clause-Clear license for development, research, prototyping, and experimentation.
• Commercial use requires a separate patent license from Zama.

Limitations & Caveats

Commercial use of the open-source code is restricted without a specific patent license from Zama. The project's technologies are patented.