Use case · Fraud & abuse
Banned on Tuesday. Back on Wednesday with a new spelling.

Repeat offenders re-register with the same name, adjusted just enough to slip exact-match blocklists. Generative tools now produce the documents to go with it.

3,000%rise in deepfake fraud attempts in a single year.Onfido Identity Fraud Report
42%of detected fraud attempts now involve AI-assisted identity manipulation.Signicat, 2024
1 charof spelling drift is all it takes to defeat an exact-match blocklist.How re-registration works
Where the check runs

At signup, against everyone you've already banned.

Every new account gets compared to the blocklist: past fraudsters, chargeback abusers, banned sellers. The attacker controls the spelling. The defense has to be spelling-proof.

Today

Blocklists that trust spelling.

  • Exact match means Katherine O'Brien and Kathryn OBrien are strangers.
  • Fuzzy rules tuned tight miss; tuned loose they flag your best customers.
  • Cross-script re-registration isn't even in the threat model.
  • Manual review queues grow linearly with signups.
With Onoma

A blocklist that survives respelling.

  • Score every signup against the blocklist; drift doesn't break the link.
  • Calibrated bands: auto-block, auto-clear, review the middle.
  • Works on names your rules have never seen, in any script.
  • Batch endpoint retro-screens the existing user base.
In the flow

Catch the re-registration at the door.

  • Signup-time scoring against the full blocklist.
  • Spelling-drift resistant: diacritics, initials, reordering, script hops.
  • Thresholds you can automate without a data-science project.
  • Evidence output for the trust & safety case file.
$ curl https://onoma.kineticpartners.org/api/v1/match \
  -H "Authorization: Bearer sk_live_..." \
  -d '{ "a": "Katherine J. O'"'"'Brien",
        "b": "Kathryn OBrien" }'

{ "score": 0.93,
  "verdict": "match",
  "evidence": {
    "script_pair": "latin",
    "normalized": { "a": "katherine j obrien",
                    "b": "kathryn obrien" } } }
Access

Make the ban stick, whatever they type next.