Paper 2025/1905

Symphony: Scalable SNARKs in the Random Oracle Model from Lattice-Based High-Arity Folding

Binyi Chen, Tsinghua University
Abstract

Folding/Accumulation schemes are a powerful tool for building scalable proof systems. However, standard approaches for leveraging folding schemes, such as IVCs or PCDs, require embedding hash functions (modeled as random oracles) into recursive circuits, introducing both security concerns and significant proving overhead. We re-envision how to use folding schemes, and introduce Symphony, the first proving paradigm that leverages folding schemes as a black box without embedding hashes in SNARK circuits. It is memory-efficient, parallelizable, and plausibly post-quantum secure, with polylogarithmic proof size and verification. The (low-memory) prover requires O(log log n) passes of the input data and its computation is dominated by committing to the input witnesses. A core component of our construction is a new lattice-based folding scheme that compresses a large number of NP-complete statements into one in a single shot. Furthermore, we design a generic compiler that converts a folding scheme into a SNARK without embedding the Fiat-Shamir circuit into proven statements.

Metadata
Available format(s)
PDF
Category
Cryptographic protocols
Publication info
A minor revision of an IACR publication in ASIACRYPT 2026
Keywords
Succinct Proof SystemsFoldingLattice-based Cryptography
Contact author(s)
by-chen @ mail tsinghua edu cn
History
2026-08-20: last of 6 revisions
2025-10-12: received
See all versions
Short URL
https://ia.cr/2025/1905
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2025/1905,
      author = {Binyi Chen},
      title = {Symphony: Scalable {SNARKs} in the Random Oracle Model from Lattice-Based High-Arity Folding},
      howpublished = {Cryptology {ePrint} Archive, Paper 2025/1905},
      year = {2025},
      url = {https://eprint.iacr.org/2025/1905}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.