Synthetic identity fraud detection is software that scores whether a new account belongs to a real person, or to a synthetic identity. A synthetic identity is an invented person, stitched together from pieces of real people's details and some made-up ones. Fraudsters use such identities to open accounts and borrow money that they never repay. This kind of fraud is largest in the United States, and so is the market for these products. So the example on this page is set there.
In the United States, the Social Security number (SSN) is the main identity number. Unlike Aadhaar, it carries no photo or biometrics. A fraudster takes a real SSN that is not being used for credit, often a child's. He adds a made-up name, a date of birth, and a new phone number and email address. He applies for a credit card. The first application is refused, but the refusal creates a credit file in the made-up name. A few months later, a card with a small limit is approved. The fraudster repays it on time for a year or more, so the bank raises the limit. Then he borrows as much as he can on several cards and disappears. This is called a bust-out. No real person complains about the account, so the bank often records the loss as an ordinary bad loan.
A synthetic identity passes simple checks, because each piece is real or correctly formed. So the software checks how the pieces fit together, using records about hundreds of millions of people. Is this SSN usually linked to a different name and age? Has this phone number or email address existed for only a few weeks? Is the same address linked to dozens of new identities? An identity graph, which links the devices, emails and addresses that belong to each real person, supplies much of this data. Device fingerprinting adds whether the same phone has been used for other applications.
Consider an application for a credit card with a $2,000 limit at a bank in Texas. The SSN is real. But the records show that it belongs to a nine-year-old, while the applicant says he is 34. The email address was created last week. The phone number has been used on four other applications this month, in other names. The software returns a score of 960 out of 999, which means the identity is very likely synthetic. The bank's credit decision engine declines the application and sends it for review.
This product is different from identity verification (IDV). IDV checks a document and a face. Synthetic identity fraud is often attempted online with no document at all, so the details themselves must be tested. In India, Aadhaar's biometric records make a fully invented person harder to create. But fraudsters still combine real PAN numbers with other people's phones and addresses.