I Deepfaked My Boss: What It Reveals About Identity Verification in the Age of AI

Disclaimer: This controlled experiment was conducted with full organizational consent and ethical oversight.Unauthorized deepfake creation is illegal and unethical.

I created a deepfake of my boss, swapping his face with mine in a live video. While not a perfect replica of our CTO, Jonathan Robins, the deepfake was realistic enough to bypass multiple face-matching and face-liveness detection systems used in common identity verification processes. This included successfully matching the deepfake to a reference image—simulating a fake ID scenario—and passing face-liveness tests against multiple detection algorithms.

What exactly is a deepfake? Simply put, it's a digital trick where artificial intelligence can digitally alter or replace someone's face in a video or image, making it look like they're doing or saying something they never did.

While this may sound impressive or even alarming, the biggest challenge for a fraudster is executing it in real time. To do so, they would need to point their camera at a computer screen, a method that modern face-liveness detection algorithms are typically designed to catch.

This experiment reinforced a crucial truth: deepfakes are advancing rapidly, but strong ID verification methods remain effective at detecting them.

Strong security measures in place, such as requiring face matching or face liveness, would prevent somebody from using a deepfake to act as a different person. Here's how:

Face Matching: A face-match check compares a selfie and the image on a government-issued ID to ensure they belong to the same person. Therefore, a deepfake alone isn't good enough, but a fake ID of the victim would need to be created as well to match the deepfake face to the face on an ID.

Face Liveness: There are two types of face liveness tests: active liveness and passive liveness. Active liveness requires a user to move their head around, while passive liveness works off a still image.

Despite the rapid advancements in deepfake technology, robust identity verification measures remain a strong defense. Face matching ensures that a fraudster must not only generate a convincing deepfake but also produce a fake ID that matches the altered face—an additional hurdle that complicates their efforts. Meanwhile, face-liveness detection helps flag deepfakes by identifying whether a face is presented from a screen rather than a live person.

The Technical Challenge: Each verification layer adds complexity to potential fraud attempts. Deepfake creators must overcome multiple sophisticated technological barriers:

●     Creating a convincing facial replica in real-time

●     Matching complex facial recognition algorithms

●     Bypassing liveness detection technologies

●     Creating supporting fraudulent documentation

However, the ultimate safeguard is multi-layered verification. A deepfake might pass face matching and even some liveness tests, but combining these checks with barcode verification on an ID creates a significantly stronger barrier against fraud.

As deepfake technology continues to evolve, so must our security measures. Businesses and individuals alike should demand strong, multi-layered authentication– including face matching, liveness detection, and ID barcode authentication to stay ahead of increasingly sophisticated fraud tactics.

About the Author

Daniel Malter is the VP of Data Science and Machine Learning at Intellicheck, where he leads a team dedicated to revolutionizing ID verification through cutting-edge technologies. His group spearheads initiatives spanning from comprehensive fraud analytics to implementing sophisticated machine learning algorithms that enhance document authentication accuracy and speed. You can find Daniel on LinkedIn at: https://www.linkedin.com/in/danmalter/