Deepfakes Can Even Trick Facial Recognition Tech

March 9, 2021

You’ve probably seen the Tom Cruise deepfakes circulating on TikTok by now.  Deepfakes like this can be very convincing, per a recent article in Venture Beat.  So, what’s going on here?

  • Deepfakes = images or videos generated by AI that replace the likeness of the person actually in the picture or video with another’s person’s likeness.

  • According to startup Deeptrace, the number of deepfakes on the web increased 330% from October 2019 to June 2020.

Tricks commercial facial recognition tech
Researchers at Sungkyunkwan University in Suwon, South Korea demonstrated that APIs from Microsoft and Amazon can be fooled with commonly used deepfake-generating methods, according to a recent paper.  Microsoft’s Azure Cognitive Services was tripped up by 78% of the deepfakes fed into it.

The study used 8,119 deepfakes of famous faces of Hollywood movie stars, singers, athletes, and politicians and found that all of the APIs were susceptible:

  • Azure Cognitive Services failed to detect deepfakes 78% of the time; and
    • the researchers successfully impersonated 94 out of 100 celebrities in one of the open source datasets in the study.

  • Amazon’s Rekognition failed to detect deepfakes 68.7% of the time;
    • Rekognition misclassified deepfakes of a celebrity as another real celebrity 40% of the time; and
    • 902 out of 3,200 deepfakes got higher confidence scores than the real image of the celebrity.

So-called anti-deepfake software often doesn’t spot the deepfakes either
The deepfake footage of Tom Cruise posted to an unverified TikTok account racked up 11 million views.  Per Vice, these deepfakes, when scanned through some of the best publicly available deepfake detection tools, avoided discovery.

This is a problem that is only increasing in prevalence and complexity without any near term tech solution in sight.

Easy peasy to share this story with your peeps

Level up your inbox with The Scroll

Get stories like this delivered to your inbox.

Business news focused on startups and tech. Get informed while being very mildly entertained.
No spam. No fluff. No nonsense. Ever.