How it works
Multi-accounting is a linking problem. Sameveil keeps an index of every identifier it has seen and asks one question for each new event: who else used this?
Strong links tie accounts into a cluster. Weak ones add weight but never form a cluster on their own, so a shared office NAT never glues strangers together.
Each rule adds points and a sentence. Points are capped at 100 and mapped to a decision. A household sharing one laptop stays under the review threshold; a farm of fifty does not. The thresholds are plain settings, tuned per operator.
Linked to an account your analyst already confirmed as fraud? That is the strongest signal we have, because a human did the work once.
{
"account_id": "player-48213",
"risk_score": 88,
"decision": "block",
"reasons": [
"email already used by 1 account(s) (dot/plus-tag variant)",
"device shared with 1 other account(s)",
"2 signups from this device in the last hour"
],
"linked_accounts": [
{ "account_id": "player-48190",
"via": ["email_hash", "device_fingerprint"] }
],
"cluster_size": 2
}
Mule accounts look clean at signup on purpose. The shared card only appears at cash-out. That is why one call at registration is not enough.
No "fraud probability 0.87". Every point has a sentence attached, and your analysts can disagree with any of them.
Sameveil never blocks anything. It answers; your platform decides. If it is slow or down, your signups keep flowing.
We do not verify documents or identities. We tell you when two "identities" are one person. Use it next to your KYC, not instead of it.