A synthesizing compiler for homomorphic encryption (HE) created by researchers at Facebook and New York and Stanford universities can translate a plain-text unencrypted codebase into encrypted code on the fly.
The compiler, named Porcupine, reportedly can accelerate HE up to 51% over heuristic-driven, hand-optimized code.
Porcupine can convert a reference of a plain-text code into HE code that performs the same computation, by internally modeling instruction noise, latency, behavior, and HE program semantics with the Quill component.
Quill lets Porcupine reason about and seek kernels that are verifiably correct while minimizing latency and noise accrual, yielding a suite that automates and optimizes the mapping and scheduling of plain text to HE code.
According to the researchers, "Porcupine abstracts away the details of constructing correct HE computation so that application designers can concentrate on other design considerations."
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA