Following a wave of scams in which fraudulent job seekers take contracts and then run off, some Fortune 500 companies have begun testing software that can spot deepfakes of real people on live video calls.
The detection technology is provided by GetReal Labs, a new company founded by Hany Farid, a professor at the University of California, Berkeley and a recognized authority on deepfakes and image and video manipulation.
GetReal Labs has developed a suite of tools to spot images, audio and videos that have been generated or manipulated, either by artificial intelligence or manual methods. The company's software can analyze faces during video calls and find clues that indicate they have been artificially generated and replaced with a real human body.
“This is not a fictitious attack, we are hearing more and more about this,” Farid said. “In some cases, it appears they are trying to infiltrate companies and get their hands on their intellectual property. In other cases, it is purely financial, they are just taking contract money.”
In 2022, the FBI issued a warning about deepfake job applicants who impersonate real people during video calls. UK-based design and engineering firm Arup lost $25 million to a deepfake scammer posing as the company's chief financial officer. Romance scammers are also employing the technique to scam unsuspecting victims out of their life savings.
Impersonating real people in live video feeds is just one example of the reality-bending tricks now possible thanks to AI. Large-scale language models can create impressive likenesses of real people in online chats, and tools like OpenAI's Sora can generate short videos. Recent advances in AI have made deepfakes more convincing and accessible. Free software makes it easy to hone your deepfake skills, and readily available AI tools can turn text prompts into lifelike photos and videos.
But impersonating people in live video is a relatively new field. Creating this type of deepfake typically involves a combination of machine learning and face-tracking algorithms to seamlessly stitch a fake face onto a real one, allowing intruders to control what their illicit likeness says and does on screen.
Farid gave WIRED a demo of GetReal Labs' technology. When shown a photo of a company's boardroom, the software analyzes the metadata associated with the image to see if it has been altered. Several major AI companies, including OpenAI, Google, and Meta, now add digital signatures to AI-generated images, providing a surefire way to verify their inauthenticity. But not all tools provide such stamps, and open-source image generators can be configured not to provide them. Metadata can also be easily manipulated.
GIF: Will Knight
