Veritus sits in front of your signup form and answers one question for every new user — real, suspicious, or fake — with a numeric score, a verdict and a list of reasons. Drop-in widget or JSON API. UK-hosted, sub-100 ms, transparent.
Veritus is a cloud-based fraud-scoring service for online signups. When a user submits a registration form on your site, Veritus enriches what they typed against multiple third-party data sources, runs the combined signals through a LightGBM machine-learning model, and returns a verdict in under 100 milliseconds.
The verdict comes in three forms so you can decide how to act on it:
allow, review, or block, computed against thresholds you controlIf your business has a signup form and someone is going to abuse it, Veritus is for you. The product is built for:
One <script> tag in your <head>, one data-veritus="signup" attribute on your form. The widget hooks the submit event, scores the user, and either lets the submit through, blocks it with a friendly message, or routes the signup to a review queue in your dashboard. Works on plain HTML, WordPress, Webflow, Wix, Squarespace — anywhere you can paste a script tag.
Developers can call POST /v1/score with a JSON payload of user details, get the same response back, and act on it server-side. Use it when you need full control of the user experience, when the signup happens outside a browser (mobile app, CLI tool, API client), or when you want to score events other than signups (login, password reset, transaction).
Every check runs through the same enrichment pipeline before the ML model sees it:
The signals feed a LightGBM gradient-boosted-trees classifier trained on a labelled corpus of confirmed real and confirmed fraud signups. The model produces a probability between 0 and 1, scaled to the 0–100 score. Two thresholds you control turn the score into a verdict:
blockreviewallowMove the thresholds in your dashboard at any time. Lower them to be stricter, raise them to be looser — you can A/B-test the impact on conversion and fraud rate side by side.
Every verdict you confirm or overturn in the dashboard becomes a training label. If you mark a blocked signup as a real customer (false positive), or release a quarantined signup that later turned out to be fraud (false negative), the system records that signal. Nightly, your accumulated feedback retrains a customer-specific layer on top of the global model, so Veritus gets better at your signup patterns over time.
Active-active deployment across two UK/EU nodes: vt1.veritus.uk in the UK on 185.65.205.151, and vt2.veritus.uk in Helsinki. PostgreSQL streaming replication and Redis master/replica keep state synchronised over mutual TLS. If either node fails, scoring continues without intervention — no DNS change, no manual cutover.
Honest caveats so you buy with eyes open:
No credit card, no trial countdown, no expiry. Every account gets 1,000 fraud checks per month for life. Upgrade when you outgrow it.