Paper 2026/2019

The Geometry of Zero Knowledge: Tight Masking and Efficient Multilinear Arguments

Xiang Wang, Guizhou University
Ji Xiang, Chongqing University
Abstract

Zero-knowledge proofs have become a fundamental tool for verifiable computation and privacy-preserving applications, driving growing interest in proof systems that are both succinct and efficient. In polynomial-based arguments, however, achieving zero knowledge introduces additional costs. Random masks must hide the evaluations and intermediate messages revealed during verification while preserving the underlying computation. Existing approaches range from protecting individual messages with cryptographic commitments to constructing small masking polynomials tailored to particular protocols. These techniques raise a more general question: how much randomness is needed when the verifier observes only a structured collection of evaluations and linear combinations? We present GoZK, a query-aware approach to zero knowledge that constructs masking spaces from the verifier’s conditional joint observations. We characterize these observations through affine images and develop low-degree masks for selected Cartesian strata, witness polynomials, and product-tree relations. In particular, two product trees that reveal eight boundary evaluations and a single root difference admit a nine-dimensional masking space, matching the dimension of their specified uniform observation interface. We integrate these constructions with a causal masking polynomial, a global sumcheck, and a pairing-based polynomial commitment scheme. The resulting protocol admits exact online simulation under the stated setup and private-coin conditions, with classical knowledge soundness established in a registered algebraic-group and random-oracle model. In a Rust implementation over BLS12-381, query-aware masking reduces the number of source-mask coefficients from 390 to 19 relative to an implemented projected arbitrary-point construction, saves 4,128 proof bytes, and improves prover time by factors of 1.40–1.67 on matched eight-column instances. Global sumcheck further reduces communication by 6,752–10,400 bytes, with a modest prover-time tradeoff.

Metadata
Available format(s)
PDF
Category
Cryptographic protocols
Publication info
Preprint.
Keywords
Zero knowledgeQuery geometrySumcheckHyperPlonkPolynomial commitments
Contact author(s)
ee wangx24 @ gzu edu cn
20240674 @ stu cqu edu cn
History
2026-11-17: last of 7 revisions
2026-09-14: received
See all versions
Short URL
https://ia.cr/2026/2019
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2026/2019,
      author = {Xiang Wang and Ji Xiang},
      title = {The Geometry of Zero Knowledge: Tight Masking and Efficient Multilinear Arguments},
      howpublished = {Cryptology {ePrint} Archive, Paper 2026/2019},
      year = {2026},
      url = {https://eprint.iacr.org/2026/2019}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.