Mpc on Ryan Orban

Piranha: A GPU Platform for Secure Computation

me@ryanorban.com (Ryan Orban) — Wed, 10 Aug 2022 00:00:00 +0000

Summary

Jean-Luc Watson, Sameer Wagh, and Raluca Ada Popa at UC Berkeley present Piranha — a GPU-accelerated platform for secure multi-party computation (MPC) that achieves 10–15× speedup over CPU-based MPC systems (USENIX Security 2022). The core insight: MPC protocols like secret sharing-based arithmetic (used in 2PC, 3PC, and 4PC settings) consist largely of matrix multiplications and element-wise operations — precisely the workloads GPUs were designed to accelerate. The obstacle has been that MPC protocols operate over custom finite fields, not the floating-point arithmetic GPU vendor libraries expect.

CrypTen: Secure Multi-Party Computation Meets Machine Learning

me@ryanorban.com (Ryan Orban) — Tue, 19 Apr 2022 00:00:00 +0000

Summary

Secure multi-party computation (MPC) lets parties compute jointly on their combined data without revealing what each party holds. The obstacle to adoption in machine learning has always been that MPC frameworks speak the language of circuit compilers and cryptographic protocols — not the language of tensors, backpropagation, and neural network modules. CrypTen from Facebook AI Research (NeurIPS 2021) fixes this by implementing an MPC computation layer that mirrors the PyTorch API almost exactly: same tensor ops, same autograd, same nn.Module structure. In principle you change one import and your code runs over encrypted data.

Secure Multiparty Computation (MPC)

me@ryanorban.com (Ryan Orban) — Tue, 19 Apr 2022 00:00:00 +0000

Summary

Yehuda Lindell (Bar-Ilan University / Unbound Tech) wrote this survey as an accessible introduction to secure multiparty computation (MPC) — the branch of cryptography that asks: how can a group of parties jointly compute a function over their private inputs, without any party learning anything beyond the output? The canonical example is a private auction: bidders can determine the winner and winning price without anyone learning others’ bids.