A new customer signs up on your app at 2 a.m. The ID is real. The selfie matches. The liveness check, where they blink and turn their head, passes without a hitch. Two weeks later, the person named on that ID calls to say they have never heard of your app.

What your system saw was the ID holder’s face, moving on cue. A fraudster was steering it with face swap software. This is face swap fraud, and it is the reason a selfie check alone no longer tells you whether a customer is who they claim to be.

This guide explains what face swap fraud is, how it gets past some biometric checks, why it matters for fintechs in Africa and which controls make it harder to pull off.

What Face Swap Fraud Is

Face swap software maps one person’s face onto another person’s live video. The output looks like the victim and moves like the fraudster. To set it up, a fraudster needs a source image of the victim, which can come from a social media profile, a leaked selfie, or a stolen ID photo.

Gartner has flagged where this leads. In 2024, it forecast that by 2026, 30% of enterprises would stop treating identity verification and authentication solutions as reliable on their own because of AI-generated deepfakes. Face swap fraud sits inside that warning.

Many verification flows were designed around a simpler threat: someone holding up a photo. A fraudster using a face swap brings a face that moves, reacts and responds on command. Whether that face gets through depends on how the fake video reaches your app.

How It Gets Past Biometric Checks

There are two routes a fake face can take into a verification system:

1. Presentation attack: The fraudster shows something to the camera, such as a printed photo, a screen replay or a mask. Liveness checks were built largely to stop this.
2. Injection attack: The fraudster skips the camera and feeds synthetic video straight into the app’s data stream, for example through virtual camera software. The system receives a video it believes came from the phone.

Gartner’s research found that presentation attacks are the most common vector, but injection attacks increased 200% in 2023. Its analyst also noted that current standards and testing for presentation attack detection do not cover injection attacks that use AI-generated deepfakes.

That gap explains why a passed liveness check is weaker evidence than it looks. A check that judges what the camera sees has little to say about video that never came from the camera. And when the face is swapped live, a prompt to blink or turn left can be answered on the spot by the fraudster, whose movements the swapped face copies.

For a fintech, the damage shows up after the account is open.

Why It Matters for African Fintechs

INTERPOL’s African Cyberthreat Assessment Report 2026 found that AI enables 55% of reported cybercrimes across Africa. It also describes criminals building synthetic identities that combine real personal data with fabricated elements, which can bypass even advanced biometric verification systems and have been used to open bank accounts, secure mobile loans, and register SIM cards under false names.

Face swap fraud adds to that pressure at the onboarding step, and the cost lands in several places:

• Fraudulent accounts: Accounts opened under a real person’s identity, ready to take loans or move money.
• Victims who find out late: The ID holder discovers an account or debt they never created, and the complaint comes to you.
• Regulatory exposure: In Nigeria, the CBN’s March 2026 baseline standards for automated AML solutions require customer identification and verification, including remote onboarding channels. A weak biometric step undermines that first function.
• Wasted compliance spend, since AML screening and monitoring downstream both run on an identity that was never real.

Biometric checks still earn their place, and the fix is to build around them.

How to Defend Against Face Swap Fraud

Gartner’s guidance points to a combination of presentation attack detection, injection attack detection and image inspection, with device identification and behavioural analytics added as extra signals. In practice, that comes down to five steps:

1. Ask your provider about injection attacks: Find out whether it detects virtual cameras, emulators and modified apps, and how it tests for them. Gartner advises choosing vendors that can show they are monitoring, classifying, and measuring these attacks.
2. Add device and behaviour signals: Look at the device, IP address, and session behaviour around each verification. These signals can disagree with what the selfie says.
3. Match strictness to risk: Use tighter liveness and face match settings for high-value products, and relax them where the stakes are lower.
4. Re-verify at risky moments: A clean onboarding does not cover the whole life of an account. Ask for a fresh liveness check before sensitive actions such as large transfers.
5. Check for repeat faces: Compare new face scans against your existing users to catch banned users coming back under new details.

Dojah’s products map onto several of these steps.

How Dojah Helps

Dojah’s liveness check asks users to complete real-time challenges such as blinking, nodding, and turning their head, and face match compares that live capture with the photo on the ID. You can set the strictness of both checks in the Fraud Check settings of EasyOnboard, so a lending product can run tighter rules than a low-value wallet.

For returning users, Easy Authentication compares a fresh liveness capture with the one stored at onboarding, which gives you a way to re-verify before critical activities. Easy Detect adds behavioural and device intelligence, analysing devices, IPs and user patterns to detect account takeovers and suspicious behaviour.

Together, these give you several checkpoints instead of one: the challenge at sign-up, the match against the ID, the signals around the session, and a fresh check before risky actions.

Ready to add layers to your onboarding? Explore Dojah’s biometric verification or talk to our team.

FAQs

What is face swap fraud?
Face swap fraud is the use of software that maps a victim’s face onto a fraudster’s live video, so the fraudster can pass a face-based identity check as someone else.

Can liveness detection stop face swap attacks?
Liveness detection helps against photos, screen replays and masks. Gartner notes that standard testing for presentation attack detection does not cover injection attacks that use deepfakes, so liveness works best alongside injection attack detection and device and behaviour signals.

How can fintechs reduce face swap fraud risk?
Ask your provider how it handles injection attacks, add device and behaviour signals, set stricter checks for high-value actions, re-verify users before sensitive actions and compare new faces against your existing users.