DRACO Design of Resilient Architectures for Computing

Encrypted Processor Framework & Accelerator

Primary Research Question: Can a modified cryptosystem support computation on encrypted data at a cost that fully homomorphic encryption cannot reach?
Encrypted Processor Framework & Accelerator research poster.
Encrypted Processor Framework & Accelerator research poster.

Fully homomorphic encryption solves the right problem and charges too much for it. Arbitrary computation over ciphertext is a genuine result, and the overhead — orders of magnitude, depending on the scheme and the circuit depth — keeps it out of most places it would be useful.

This project takes the opposite starting point. Rather than accepting FHE’s generality and attacking its cost, we modify a well-understood cryptosystem to support the operations that matter, and build the accelerator that makes those operations fast. The target is deliberately in between: stronger than computing on plaintext, and cheap enough to be a real option where FHE is not.

Two halves, and neither works alone:

The work is a sibling to the lab’s Side-Channel-Resistant Computer Architecture project, which arrives at overlapping territory from the architecture side.

Open questions