Science and TechArtificial Intelligence


How to tell the difference between a deepfake video and a real one

There are some things to pay attention to when identifying a deepfake video, but artificial intelligence is making detection more difficult.
This image made from a fake video featuring former President Barack Obama shows facial mapping elements used in deepfakes.
Posted at 4:25 PM, May 16, 2023

Do you know how to spot a deepfake video?

Deepfake technology first appeared in November 2017, according to the Organization for Social Media Safety.

You may remember a deepfake video of Ukrainian President Volodymyr Zelensky that circulated back in 2022. It showed the leader asking soldiers to lay down their weapons. He later addressed the fabricated, false statements.

But as technology advances, fabricated videos are starting to look more and more like real ones, which could cause a lot of confusion.

James Turgal, the vice president of cyber risk at Optiv and a former FBI cybersecurity expert, wrote to us in an email interview, "Deep learning technology is a kind of machine learning that applies neural net simulation to extremely large data sets. The advent of faster and smarter artificial intelligence algorithms takes in this mass amount of data — video in this case — and the algorithm then learns what a particular face looks like at different angles in order to transport the face onto a target, just as if it were a mask."

Fake people, real threat: How AI is being used to spread propaganda

Fake people, real threat: How AI is being used to spread propaganda

A new report highlights how life-like, AI-generated influencers are being used to promote Chinese propaganda and discredit American ideals online.


So how can you tell the difference?

Turgal said there are a few things you can watch out for including unnatural eye movement, lack of blinking, pixelation, or a misaligned background.

However, with the advancement of artificial intelligence, some of these identifiers have become more difficult to spot. Turgal wrote to us in an email, "With the use of Generative AI, the ability to spot deep fakes will only grow more difficult. There is also a growing push to build tools to help detect synthetic people and media."

He noted that the job of policing these videos is up to the algorithms or certain platforms, and ultimately, it will be left up to algorithms and humans to deny the deepfakes.

Experts at the Massachusetts Institute of Technology also suggest paying attention to the cheeks and forehead, the glare on glasses, facial moles, and lip movements.

Paying attention to where the video or information is coming from is also important.

Why Deepfake Pornography Is So Hard To Stop (In The Loop)

Why Deepfake Pornography Is So Hard To Stop (In The Loop)

Christian Bryant explains why cracking down on deepfake revenge porn and celebrity deepfakes has been difficult legally.