Reproduce “HEIR: A Universal Compiler for Homomorphic Encryption” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Asra Ali, Jaeho Choi, Bryant Gipson, Shruthi Gorantala, Jeremy Kun, Wouter Legiest, Lawrence Lim, Alexander Viand and 2 more
Researchers from Google and several academic institutions developed HEIR, a universal compiler framework for homomorphic encryption built on MLIR. This framework unifies mainstream HE schemes and their diverse optimization techniques, integrating with various software libraries and hardware accelerators, and is gaining traction as a standardization candidate for the field.
Key insights
- MLIR provides a suitable and flexible foundation for constructing a universal HE compiler capable of abstracting and integrating diverse HE schemes, libraries, and hardware targets.
- A layered, dialect-based compiler architecture effectively encapsulates and manages the wide array of previously disparate HE optimization techniques, streamlining development and enabling systematic comparison.
- Automating complex, manually intensive HE optimizations like parameter selection, noise modeling, and data layout within a unified compiler framework is feasible and reduces the need for highly specialized cryptographic expertise.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.