veritus.ukAll articlesHome

How to Spot Fake Signups: The Warning Signs Hiding in Your Registration Data

Updated 16 November 2026

If you run a SaaS free trial, an online shop with promo codes or a community forum, you have probably looked at a list of new accounts and thought: how many of these are real? Fake signups rarely announce themselves. They often look perfectly ordinary until you know where to look.

This guide covers the warning signs that tend to show up in registration data, why each one on its own can mislead you, and how to combine them into decisions you can stand behind.

Why fake signups are worth catching early

A fake account is cheapest to deal with before it exists. Once it is in your database it can claim a free trial, use up compute, redeem a welcome discount, post spam or skew your growth figures. Cleaning up afterwards takes staff time, and some damage, such as an abused promo code, cannot be undone.

That is why the signup form is the natural checkpoint. Everything a person types there, plus how and where they typed it, carries clues.

The warning signs, field by field

The email address

The phone number

The IP address and location

The password

A password that appears in public breach lists suggests either poor habits or a credential lifted from a leak. Checking this does not require seeing the password itself: a technique called k-anonymity sends only a short fragment of a hash, so the real password never leaves the browser.

How the form was filled in

Patterns across many signups

Some fraud only shows up in aggregate: bursts of signups from the same IP range, near-identical email patterns or the same device repeating. Velocity and repeat-signup checks look for exactly this.

Why no single signal is proof

The biggest mistake is to block on one clue. Each warning sign has an innocent explanation, and treating them as verdicts will turn away real customers.

Warning signCould mean fraudCould also be
VPN connectionHiding true locationA privacy-conscious customer or a corporate network
New email addressCreated just for abuseSomeone who set up a fresh inbox for your service
Country mismatchStolen or invented detailsA traveller, or an expat with a home-country phone
VoIP numberBulk-bought verification numberA business using an internet phone system

What separates fraud from coincidence is the combination. A three-day-old email, a VoIP number, a Tor exit node, a breached password and a UK form submitted from a Russian IP together tell a very different story from any one of them alone.

Turning signals into a decision

A sensible approach uses three outcomes rather than two:

The review band is what protects your conversion rate. It lets you be cautious without being harsh, and your review decisions become the best evidence you have about what fraud looks like on your site. Keeping a human review path also matters wherever an automated decision could significantly affect a person.

A quick audit you can run this week

  1. Export your last month of signups with email, phone, country and IP if you log it.
  2. Sort by email domain and look for disposable or nonsense domains.
  3. Group by IP or IP range and look for bursts.
  4. Compare stated country with phone country code.
  5. Check how many accounts never verified their email or never logged in again.
  6. Note which combinations appear most often among the accounts you are confident are fake.

Even a rough pass usually reveals a pattern, and tells you whether you need something automated.

Where Veritus fits

Doing this by hand works for a handful of signups. At any real volume it needs automating, which is what Veritus does at the moment of signup. Each check runs seven enrichment signals — email validity and age, phone validity and line type, address validation, IP and geolocation, breach exposure, device and behaviour, and cross-field consistency — then a LightGBM machine-learning model weighs them together and returns a 0–100 score in under 100 milliseconds.

Every score comes with human-readable reason codes such as ip.tor_exit or geo.mismatch (GB/RU), so you can see why a decision was made. You set the review and block thresholds (by default 50 and 85), cross-consistency mismatches are flagged rather than auto-blocked, and the model learns from the verdicts you confirm or overturn. You can read the full signal list on the features page or follow a signup step by step in how it works.

Veritus is UK-hosted, and you can add it with a single drop-in widget script tag or through a REST API. Worth knowing: behavioural signals such as time-to-fill and paste-versus-type are collected by the widget only.

Try it on your own signups

The easiest way to see what is really signing up to your site is to score it. The Free plan gives you 1,000 checks a month with all seven signals, no card required, and paid tiers start at £29 a month when you need more volume, a customer-specific model and a review queue. See the options on the Veritus pricing page.