Alt of whom, not just suspicious.
Get the linked accounts back in your own user IDs, with the evidence on every edge named and weighted.
One account, four links.
Already in production
Publishing platforms, fintech and national charities across Europe already run TrustSig on their forms.
PublitPrint-on-demand and digital book distribution for publishers.Publishing · Sweden
LEI SystemOfficial Legal Entity Identifier registration agent.Fintech · Estonia
Disability Sport WalesNational charity for inclusive and disability sport.Charity · Wales
Four ways two accounts are one person
Each link carries its own confidence, and the response names which kind fired, so a reviewer can check the evidence behind it.
Gate on it, or investigate with it
You run one inline lookup at signup, and ask for the full graph only when a human is reviewing the account.
const v = await pro.verify(token, { user_id: user.id, traits: { email: user.email }, include: ["alt"],});
if (v.alt?.is_alt) { const [who] = v.alt.accounts; return rejectDuplicateEntry(who.user_id);}// Investigating one user: adds linked and spoofed devices.const deep = await pro.getAlts(user.id);
// One device, N of your users.const seen = await pro.getEntityAccounts(v.canonical_entity_id);No alts and could not tell are different answers
A linking system that never admits doubt will eventually tie two strangers on office Wi-Fi together. A weak answer comes back with its reason.
- Cohort-grade hardware, where many identical phones share a fingerprint, suppresses same-device matches rather than inventing them.
- A device the resolver has not finished resolving is reported as unresolved, never as clean.
- An identifier the visitor merely typed cannot read the alt graph, so nobody enumerates an account they do not own.
- Confirm or reject a link and the correction feeds your own tuning rather than a global model.
- Trust
- 84 → 21
- Sessions
- 412
Find the users you are already paying for twice.
Talk to an expertAlt accounts, answered
Four link kinds, each with its own confidence. same_device is the same hardware under another of your users, and linked_device is a device the identity graph tied to theirs. email_variant is one mailbox normalised, so a.b+promo@gmail.com matches ab@gmail.com, and spoofed_device is a match that survived a changed fingerprint.
Your own user IDs come back, with the link kind, the confidence and when each account was first and last seen. The answer reads: u_4821 is an alt of u_9907, by same_device, at confidence 88.
Both. Add include: ['alt'] to the verify call you already make and the multi-account verdict comes back inline. The deeper per-user lookup also covers accounts on linked and spoofed devices, which is what you want when investigating.
Shared-IP evidence alone carries low confidence for that reason. Cohort hardware, where many devices share one fingerprint, comes back as degraded rather than matched, which means could not tell. Every link ships its evidence, so borderline cases stay inspectable.
The device view lists every account ever seen on one piece of hardware, newest first, with per-account event counts. A phone farming accounts has no linked devices, so the alt lookup stays quiet and the device view fills up.
Ask the same platform a different question
Every page below runs on the same telemetry, the same device identity and the same detection catalog.
See how many of your users are one person
Tell us where the duplicates hurt: promos, trials, referrals or bans. We will reply by email to set up access for your team.
- EU-hosted and GDPR-native
- Cookie-free device identity
- Training mode before anything enforces