The Darker Side of Fake News: Deepfakes Explained

If the technology existed for a sinister hacker to manipulate any video of a famous person to say whatever the hacker wanted, how could you ever trust the authenticity of a video ever again? What if we told you not only does this technology exist, but it’s already in widespread use? This isn’t science fiction. This is very real, and tech giants are starting to take action. Earlier this month, Facebook announced that it will ban all “deepfake” videos from its platform. This stance against media manipulation made big headlines after the social network’s announcement. Why all the fuss, you may ask. What even is a deepfake video and why should you care?

Deepfakes 101

Deepfakes are videos altered using cutting-edge artificial intelligence (AI) to distort reality. The name is actually a portmanteau of “deep learning” (artificial intelligence) and “fake” videos. This growing technology analyzes real images to generate fake ones. Deepfakes are part of the rising spread of disinformation, and social media platforms have been struggling with how to deal with it, including Facebook. Deepfake videos started in 2017 by a Reddit user who swapped the faces of female celebrities with the faces of porn stars. The process takes only a few steps. You must first run thousands of photos or videos of the faces you want to manipulate through an AI algorithm that finds and learns similarities between the two faces. A second tool, called the decoder, then reconstructs the face of the victim to speak and make facial changes based on the hacker’s desire. Want to see it in action? Here are some incredibly good deepfakes:

Jim Carrey in The Shining

https://www.youtube.com/watch?v=-ZRUZzZPGto

Bill Hader as Tom Cruise and Seth Rogen

https://www.youtube.com/watch?v=VWrhRBb-1Ig

More Deepfakes

https://youtu.be/ZJrffEfCMrs

Detecting Deepfakes

Even experts struggle to detect advanced deepfake videos. Proper detection will only get more difficult as technology improves. For example, in 2018 researchers discovered that most deepfakes didn’t show the victim blinking. But soon after the research was published, the deepfakes upgraded and included normal blinking. Governments, universities, and tech firms are all now funding research to detect deepfake videos. Recently the first Deepfake Detection Challenge began, an effort backed by Microsoft, Facebook, and Amazon. The Challenge will include research teams around the world who are competing for supremacy in the deepfake detection game. Time will tell how effective some of the measures taken to prevent harmful deep fake videos will be. What are your thoughts about deepfake videos? Are you concerned or do you think it’s not an issue worth stamping out? Let us know on Facebook or LinkedIn.