Design of a hardware hash engine for accelerating STARKs on resource constrained devices

Abstract

In recent years, zero-knowledge proofs have made great strides in efficiency, on two fronts: minimizing the finite field size without sacrificing security and allowing more constructs in their intermediate representations, such as look-ups. These have lead to novel, efficiently aritmetizable, hash functions, which are the backbone of modern proof systems. We take advantage of these improvements and present a resource constrained hardware design for the Monolith hash function, targeted torwards embedded, edge IoT devices. The hardware implementation demonstrates a significant performance advantage over the software-only solution on the Zynq SoC, achieving a 22× speed-up while consuming only a few hundred milliwatts.