TenSEAL

A library for doing homomorphic encryption operations on tensors, built on Microsoft SEAL.

TenSEAL bridges the gap between high-level Python tensor operations and the complex C++ backend of Microsoft SEAL. It enables encrypted data processing for NumPy and PyTorch workflows, allowing developers to perform vector additions and matrix multiplications on ciphertext without decrypting it first. By implementing schemes like BFV and CKKS, TenSEAL secures sensitive machine learning tasks (such as private inference) while maintaining a familiar API for data scientists. It is a core component of the OpenMined ecosystem, designed to scale privacy-preserving computations across distributed environments.

https://github.com/OpenMined/TenSEAL