Dataset
The AttestSeal trust dataset.
One million+ web domains, scored with the current attestseal-v1.5.1-weights model, each with a signed evidence bundle that can be verified offline against our published DID.
1,001,058Domains scored
3,680,000+Raw signal records
39,187PROCEED calls
57.3Average trust score
Use it now via API
The full dataset is queryable today. Every check is signed and verifiable:
GET https://api.attestseal.com/v1/check/{domain}
See the API documentation for the response shape, or the field-by-field walkthrough.
Downloadable snapshot
Coming soon
We are preparing a public CC-BY-4.0 snapshot of the dataset for HuggingFace and a GitHub Release. The snapshot will include the full scored registry as CSV + JSON with a SHA-256 manifest. Subscribe to the blog or watch the GitHub repo for the publication announcement.
What is in each row
- Identity: domain, checkedAt, expiresAt, scoringModel
- Call: trustScore, recommendation (PROCEED / CAUTION / DENY), confidence, cautionReason, brandTier, assuranceBasis
- Signals: per-family scores and evidence (domain age, SSL, DNS, content, reputation, identity)
- Flags: WELL_KNOWN_BRAND, NO_SSL, MALWARE_DETECTED, etc.
- Provenance: Ed25519 signature, signatureKeyId, issuer (did:web:attestseal.com)
Related
- Scoring methodology — how the calls are computed
- The compositional brand anchor — why long-term Tranco rank + clean reputation + valid SSL is a defensible trust floor
- Public legitimacy vs merchant trust — the assuranceBasis taxonomy