Welcome to the resource topic for 2024/1918
Title:
Orion’s Ascent: Accelerating Hash-Based Zero Knowledge Proof on Hardware Platforms
Authors: Florian Hirner, Florian Krieger, Constantin Piber, Sujoy Sinha Roy
Abstract:Zero-knowledge proofs (ZKPs) are cryptographic protocols that enable one party to prove the validity of a statement without revealing the underlying data. Such proofs have applications in privacy-preserving technologies and verifiable computations. However, slow proof generation poses a significant challenge in the wide-scale adoption of ZKP. Orion is a recent ZKP scheme with linear prover time. It leverages coding theory, expander graphs, and Merkle hash trees to improve computational efficiency. However, the polynomial commitment phase in Orion is yet a primary performance bottleneck due to the memory-intensive nature of expander graph-based encoding and the data-heavy hashing required for Merkle Tree generation.
This work introduces several algorithmic and hardware-level optimizations aimed at accelerating Orion’s commitment phase. We replace the recursive encoding construction with an iterative approach and propose novel expander graph strategies optimized for hardware to enable more parallelism and reduce off-chip memory access. Additionally, we implement an on-the-fly expander graph generation technique, reducing memory usage by gigabytes. Further optimizations in Merkle Tree generation reduce the cost of SHA3 hashing, resulting in significant speedups of the polynomial commitment phase. Our FPGA implementation heavily optimizes access to the off-chip high-bandwidth memory (HBM) utilizing memory-efficient computational strategies. The accelerator demonstrates speedups of up to 381$\times$ for linear encoding and up to 2,390$\times$ for the hashing operations over a software implementation on a high-end CPU. In the context of real-world applications, such as zero-knowledge proof-of-training of deep neural networks (DNNs), our techniques show up to 241$\times$ speed up for the polynomial commitment.
ePrint: https://eprint.iacr.org/2024/1918
See all topics related to this paper.
Feel free to post resources that are related to this paper below.
Example resources include: implementations, explanation materials, talks, slides, links to previous discussions on other websites.
For more information, see the rules for Resource Topics .