Hardware Acceleration For Multi-Scalar Multiplication In Zero-Knowledge Proofs
Engineering
CSE 195 Independent Thesis Research
As more and more aspects of life begin to rely on the internet, secure communication
and secure computation are becoming extremely important. Zero-Knowledge Proofs
are cryptographic protocols that allow one party to prove a statement to another party
in a way that doesn’t allow for the extrapolation of any additional information that
the prover does not explicitly want the verifier to know. These protocols have many
applications in blockchains, verifiable outsourcing of work, digital signatures, or proving
statements about private pieces of data. These proofs, while effective, put enormous
pressure on CPU’s as the generation of these proofs is rather non-trivial.
One computationally demanding phase of Zero-Knowledge Proof generation
is multi-scalar multiplication (MSM) on elliptic curves. In this work, we accelerate
MSM with a custom hardware accelerator designed in the Chisel Hardware Description
Language. This higher level HDL allows for easy parameterization so we can easily
explore the design space. Our design shows 4.5-13x speedups over our functional model
of Multi-Scalar Multiplication. We are also able to scale up the parallelism of the design
to observe a reduction in cycle count. With custom hardware to handle this time-
consuming operation, these zero-knowledge proofs can become more feasible for widescale adoption.
Adoption of these protocols would increase privacy between parties and
help foster collaboration between parties that would otherwise not trust each other.
