The Scale of the Problem
Synthetic identity fraud is not a future threat for Nigerian financial institutions. It is happening now, and the losses are significant. Unlike traditional identity theft where a criminal steals someone's real identity, synthetic fraud involves combining real and fabricated details to create a person who does not exist. A real Bank Verification Number might be paired with a fake name, a manufactured address, and a phone number registered to a SIM that was bought in a bus park. The result is a borrower or account holder who passes most basic checks but was never a real person at all.
Nigerian banks and fintechs lose billions of naira annually to this form of fraud, and much of it goes undetected for months because synthetic identities are designed to behave well at first. They build credit histories, maintain low transaction volumes, and avoid triggering obvious red flags. By the time the institution realizes something is wrong, the synthetic identity has already cashed out.
How Synthetic Identities Are Built
The process is more organized than most compliance teams realize. Fraud rings operating in Nigeria often use leaked BVN databases, which are unfortunately not difficult to obtain. These BVNs are matched with clean phone numbers, fabricated addresses in legitimate streets, and email addresses that look plausible. The assembled package is then submitted through onboarding flows that rely heavily on document verification without deeper behavioral analysis.
Where document verification is stronger, fraudsters turn to corrupted onboarding agents who can push through applications with minimal scrutiny. The agent network that Nigerian fintechs and mobile money operators depend on for customer acquisition is also a vector for synthetic identity injection.
Why Standard KYC Fails to Catch It
Standard KYC processes verify that the documents match the details submitted. They do not verify that the person submitting those details is the same person the BVN belongs to. This is the gap that synthetic identity fraud exploits. As discussed in our analysis of continuous risk monitoring, a one-time check at onboarding is not sufficient to protect an institution over the life of a customer relationship.
Liveness detection and biometric verification help, but they are not foolproof when the BVN and supporting documents are authentic, just not belonging to the person presenting them. Institutions need layered controls that go beyond the onboarding moment.
Detection Strategies That Work
The most effective approach combines velocity checks, network analysis, and behavioral monitoring. Velocity checks flag when the same device, IP address, or phone number is used to open multiple accounts. Network analysis identifies clusters of accounts that share uncommon attributes like similar onboarding patterns, overlapping contact details, or accounts that refer each other.
Behavioral monitoring is where machine learning adds real value. Platforms like Remllo's WatchTower can identify transaction patterns that deviate from what a legitimate customer in a given segment would produce, even when individual transactions look clean.
The combination of network graph analysis and behavioral baselines makes it possible to surface synthetic identities before they cause significant loss, rather than discovering them during a write-off review.
What Institutions Should Do Now
The first step is an honest assessment of your current onboarding controls. If your verification process stops at document matching and a liveness check, you have gaps that sophisticated fraud rings know how to exploit. The second step is introducing post-onboarding monitoring that does not rely on the assumption that a clean application means a clean customer.
Finally, institutions should invest in data sharing mechanisms with peers. Synthetic identity fraud tends to move across institutions as fraudsters test where controls are weakest. Industry-level data sharing, even at an aggregated level, significantly raises the cost of running synthetic identity operations at scale.
