onshore
Products / Core

Finds a BVN even when nobody labelled it as one.

People paste account numbers into chats, tickets and log messages. The detection engine reads free text and JSON and finds the identifiers, even when they sit among other digits and words.

DetectionProtecting
Transfer of N250,000 to 0123456789 failed. My BVN na 22123456789
BVN0.97Tokenized
NUBAN0.91Tokenized
Amount0.93Tokenized
The problem

Field rules only protect the fields you remembered. The real exposure is the BVN inside an error message or a customer's complaint.

How it works
01

Read the text

Text and JSON go through pattern checks, validators and the words around a value.

02

Score each match

Every match gets a type and a confidence score, so weak matches can be handled differently.

03

Act on it

Tokenize, mask, drop or simply count, depending on your rule.

In code

Clean any text

const out = await vault.detect(
  "Transfer of N250,000 to 0123456789 failed. My BVN na 22123456789",
  { tokenize: true, vault: "v_live_ng01" },
);

// "Transfer of tok_amt_9Lk2 to tok_nuban_Qa81LmZ3Xe0d failed.
//  My BVN na tok_bvn_7Hq2xP9LmA4c"
What's included

Built for Nigeria

BVN, NIN, NUBAN, phone numbers, cards, emails, names and naira amounts.

Your own data types

Add loan IDs, staff numbers or VIP lists through keyword, pattern, after-keyword, value list or JSON path rules.

Try before you enforce

New rules start in Watch only. See what they would catch in real traffic before they change anything.

Approvals

A second person can be required before a rule starts protecting.

Safe by design

Linear-time matching and size limits stop hostile input from slowing the engine down.

Never double-tokenizes

The engine recognises its own tokens and leaves them alone.

Who uses this

Start with one leak.

Tell us which flow worries you most. We'll show you what it would take to keep it onshore.