Google's C++ to FHE Transpiler by Shruthi Gorantala & Rob Springer 26.08.2021

Abstract

In this talk, Shruthi and Rob present the Fully Homomorphic Encryption Transpiler recently released by Google. The FHE transpiler automatically converts high-level [C++] code that operates on unencrypted data into a form that performs the same computation on encrypted data.

They discuss the inner workings of the transpiler: how input C++ is converted via the XLS hardware synthesis toolchain into the gate-level operations used by TFHE. They then describe how the transpiler removes developer barriers and will accelerate research and drive greater FHE adoption.

Finally, they demo the transpiler and demonstrate some of the new use cases it enables.

About the Speakers

Shruthi Gorantala is a software engineer at Google with a focus on privacy infrastructure and private computing. She received her masters in computer science from University of Pennsylvania and her interests include machine learning and fully homomorphic encryption.

Rob Springer is a software engineer at Google, working on the XLS hardware synthesis toolchain. Previously, he worked on low-level software and firmware for Google’s TPU and VPU accelerator stacks and on other large-scale distributed systems and accelerators. Rob earned his bachelor’s degree in computer science from Emory University and his master’s degree in the same from the University of Georgia.

Links