HN Showcase
Back to browse
Sign a dataset so anyone can verify it offline, without trusting you

Sign a dataset so anyone can verify it offline, without trusting you

by christ_dev·Aug 4, 2026·2 points·0 comments
Visit ProjectView on HN

AI Analysis

●●●BangerBig BrainSolve My ProblemZero to One

Names the exact row changed in a dataset, not just that the file hash mismatched.

Strengths
  • •Granular verification names specific modified entries instead of failing the whole file.
  • •Public transparency log decouples proof from trusting the data publisher's infrastructure.
  • •Encrypts specific columns while keeping the rest of the container queryable without keys.
Weaknesses
  • •Requires adopting a custom .surx container format rather than signing existing Parquet files.
  • •Relies on a new transparency log service that needs long-term availability guarantees.
Category
Security
Target Audience

ML engineers, data scientists, and compliance teams managing training data

Similar To

Sigstore · Docker Content Trust · Notary

Similar Projects

Security●●Solid

AERF, signed receipts for AI agent actions

Cosign for agent evidence with compliance mappings, but still draft spec.

Big BrainBold Bet
keertahacker
212mo ago
Developer Tools●●●Banger

BootProof – the run button that can't lie (boots any repo, signs proof)

Self-verifying HTML receipts with ed25519 signatures prove your repo actually boots.

WizardryBig BrainSolve My Problem
bootproof
201mo ago
AI/ML●●●Banger

AvatarBook – Verifiable AI agent workflows

Ed25519-signed execution traces solve agent verification before it becomes a crisis.

Zero to OneBig Brain
noritaka88
103mo ago