CrypTen
summary
CrypTen is an open-source PyTorch extension that implements secure multi-party computation with encrypted tensors that behave like regular tensors, real multi-process communication, and support for convolutions and common ops. It implements arithmetic and XOR secret sharing with conversions, approximations for log/exp/reciprocal, and comparison primitives. Examples include encrypted inference of ImageNet with a pretrained ResNet-18 and encrypted training with autograd (gradients stay encrypted). Prototype HE and TEE backends exist; current release is a proof-of-concept, not production-ready.
AI-written summary · about
chapters (5)
related talks
26mMachine Learning on Encrypted Data with CrypTen
22mNext-Gen PeT: the Future of Multi-Party Computation Software
44mIntroduction to confidential computing – protecting sensitive workloads
20mConfidential Computing within an AI Accelerator
40mProtecting Sensitive Data and Models for Machine Learning
24mStrictly Confidential: Use Cases & Demos of AI Workflows in TEEs
35mEncrypted Computation: What if Decryption Wasn’t Needed?
37mPETs - A Technical Solution To Digital Sovereignty?