Building a compiler and processor for fast processing of encrypted data
summary
Claims a toolchain that takes C, compiles it to a compact FHE instruction set, and runs on a parasol virtual processor with a compiler from C to FHE. Uses Sunscreen’s Taurus‑based primitives, 128‑bit security parameters (error ~2^-40), and a cacheable compute key; achieves ~610 ms for a chi‑squared benchmark on an AWS C7G6x large, <100 ms compile time, and ~0.25–0.5 KB binaries. Demos (dark‑pool matching, on‑chain image convolution via an offchain HPU) work end‑to‑end, subject to block‑time latency.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:25Traditional data processing risks
- 5:34Fully homomorphic encryption
- 9:23FHE challenges today
- 11:49Sunscreen FHE stack
- 13:34First FHE program in C
- 20:16Live encrypted addition
- 32:48Advanced FHE features
- 40:08Private trading demo
- 45:04Image convolution demo
- 52:28Performance benchmarks
- 54:33Q&A
related talks
26mBuilding Private And Secure Applications Easily Using The Sunscreen FHE Compiler&Rust
17mFHE Rollups and Co-Processor in Action
36mEverybody is talking about Confidential Computing - the minimum Devs should know
40mConfidential Computing and Privacy-Enhancing Technologies - The Landscape
41mPanel: A New Root Of Trust: Changing Computing With TEE
18mHardware-Software Codesign for Mitigating Spectre
58mFast Cryptographic Code via Partial Evaluation
20mConfidential Computing within an AI Accelerator