Paper 2026/2946

BOCCHI: Compact Blind SNARK via Malformation-Tolerant Ciphertext Aggregation

Ye Yu, National University of Singapore
Kexi Huang, National University of Singapore
Abstract

Blind SNARKs combine fully homomorphic encryption (FHE) with cryptographic proofs to enable verifiable computation over encrypted data, and can substantially outperform SNARK-over-FHE approaches in efficiency. However, ciphertext openings in existing constructions can inflate proofs to tens or hundreds of megabytes. We present BOCCHI, a compact and efficient blind SNARK that addresses this bottleneck by committing to ciphertexts natively and opening them through a few aggregates. To establish soundness even for malformed ciphertexts, we introduce canonical binding and a lightweight noise filter that achieves it with optimal error, tools that may be of independent interest. We further formulate this transformation as a Commit–Aggregate–Check (CAC) compiler and show its effectiveness on two systems with different PIOPs (GKR and Fractal). At $2^{22}$ constraints, BOCCHI produces a 6.57 MB proof in 115.3 seconds, 9.3× smaller and 2.54× faster than Phalanx. Its proof is the smallest among five baselines. Our implementation is available at https://anonymous.4open.science/r/BOCCHI-Artifacts.

Metadata
Available format(s)
PDF
Category
Cryptographic protocols
Publication info
Preprint.
Keywords
blind SNARKfully homomorphic encryptionverifiable computationproof delegation
Contact author(s)
ye yu @ u nus edu
kexi huang @ u nus edu
History
2026-11-22: approved
2026-11-19: received
See all versions
Short URL
https://ia.cr/2026/2946
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2026/2946,
      author = {Ye Yu and Kexi Huang},
      title = {{BOCCHI}: Compact Blind {SNARK} via Malformation-Tolerant Ciphertext Aggregation},
      howpublished = {Cryptology {ePrint} Archive, Paper 2026/2946},
      year = {2026},
      url = {https://eprint.iacr.org/2026/2946}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.