← Back to all articles

How Companies Detect Fake Addresses: Techniques Used by Banks, Insurers, and E-Commerce

Fraudsters rarely hand-type a random string and call it an address anymore. Modern fake addresses are synthetic: they look plausible, may even exist, but are not the user's own. Detecting them is a layered problem.

Postal database validation

The first and cheapest layer is a lookup against a postal reference database. It catches typos and truly nonexistent streets instantly. But a synthetic address that copies a real, vacant property passes this layer — which is why it cannot be the only check.

Behavioral and linkage signals

How the address was entered matters as much as what was entered. Pasting an address, correcting the ZIP multiple times, or using a form-filler all send signals. Linkage analysis goes further: has this address appeared on dozens of unrelated accounts, or is the same IP applying with different addresses?

Machine learning and velocity

Models score addresses on features invisible to a human reviewer: distance to shipping address, age of the address in the database, and consistency with past behavior. A brand-new account shipping a high-value item to an address never seen before scores higher risk than an established account shipping to its usual place.

For QA and data teams, the lesson is simple: generate addresses that are structurally real, so your detection logic is tested against believable inputs, not strawmen.