University of Illinois Urbana-Champaign

ZKPerf: Performance benchmarking and Analysis of Zero-knowledge Frameworks and Virtual Machines

Alluri, Lakshmi Saivenkata Siddhartha Varma

This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.

Permalink

https://hdl.handle.net/2142/129573

Description

Title
ZKPerf: Performance benchmarking and Analysis of Zero-knowledge Frameworks and Virtual Machines
Author(s)
Alluri, Lakshmi Saivenkata Siddhartha Varma
Issue Date
2025-04-24
Director of Research (if dissertation) or Advisor (if thesis)
Kang, Daniel
Department of Study
Siebel School Comp & Data Sci
Discipline
Computer Science
Degree Granting Institution
University of Illinois Urbana-Champaign
Degree Name
M.S.
Degree Level
Thesis
Keyword(s)
  • Cryptography
  • ZK-SNARKs
  • Benchmarking
Abstract
Recent advancements in zero-knowledge succinct non-interactive arguments of knowledge (ZK-SNARKs) have led to the development of high-performing, accessible, and well-packaged tools for building ZK-SNARKs. This tooling enables developers to describe arbitrary computations from high-level definitions, supporting applications that range from ensuring the integrity of machine learning inferences to scaling blockchain transactions. Internally, these tools are built on various proof systems, commitment schemes, and arithmetization techniques, offering distinct performance characteristics that depend on the task. However, existing benchmarks are often granular and compare only a limited set of lower-level primitives and are limited in coverage of available tools. In this work, we present a comprehensive benchmark for evaluating the end-to-end performance of the ZK-SNARK tooling. We apply this benchmark across four popular frameworks and two virtual machines (VMs), comparing performance on four tasks: Elliptic Curve Digital Signature Algorithm (ECDSA) verification, Merkle Tree membership verification, an MNIST Convolutional Neural Network (CNN), and Meta's Deep Learning Recommendation Model (DLRM). Our results reveal that different provers excel in different areas, with the fastest prover for ECDSA proving MNIST CNN 6.7x slower than the fastest prover for MNIST CNN. We also investigate how implementing lower-level primitives such as Multi-Scalar Multiplications (MSMs) and Fast Fourier Transforms (FFTs) in the frameworks impacts the performance of the same protocol, observing a 7x difference in performance. Additionally, we observe up to a 10x variation in lookup performance across frameworks implementing the same lookup protocol as the number of queries increases for a fixed constant table size. We also compare the trade-offs between expressiveness and performance in ZK-SNARK frameworks and virtual machines (VMs). Finally, we open-source our implementation to serve as a foundation for the community in standardizing benchmarking for ZK-SNARK tooling.
Graduation Semester
2025-05
Type of Resource
Thesis
Handle URL
https://hdl.handle.net/2142/129573
Copyright and License Information
Copyright 2025 Lakshmi Saivenkata Siddhartha Varma Alluri

Owning Collections

Graduate Dissertations and Theses at Illinois PRIMARY
Graduate Theses and Dissertations at Illinois