You may use LinkedIn to congratulate a friend on their work anniversary, but spies use it to find new potential sources, and they may be using AI-generated profiles to help them do it.
A new report from the Associated Press shows possible evidence of a fake profile being used to lure in potential contacts on LinkedIn. Katie Jones’s profile pic is pretty standard stuff for LinkedIn. The photo shows a nice-looking, clean-cut white lady with blue-ish eyes and light brown hair and an inscrutable expression on her face, like a LinkedIn Mona Lisa. The image is pretty generic, which is the point, as it seamlessly blends in with all the other nice-looking, clean-cut LinkedIn profiles eager to connect.
However, “Katie Jones” may be special, because according to the AP there’s a good deal of evidence that it was created by AI.
Several AI experts told the AP that the image of “Katie Jones” was definitely created using machine-learning techniques. A few of the subtle-yet-telltale signs that Katie Jones was created on a computer somewhere include a slightly asymmetrical face, an earring that appears to have melted off, an indistinct background, blurry boundaries between hair and ear, and streaks on her cheek that are either blur marks or a sign that she has been crying in the bathroom at work.
Why would anyone bother using machine-learning techniques to create a profile pic of an average white lady? Because it works. According to the AP, “Katie Jones” used LinkedIn to connect with policy experts and government figures in Washington, including a senator’s aide, a deputy assistant secretary of state, and Paul Winfree, an economist currently being considered for a seat on the Federal Reserve. Jones requested a connection, and they all accepted, because what the heck, it’s LinkedIn. As Winfree told the AP: “I literally accept every friend request that I get.”
Why LinkedIn? Because as William Evanina, director of the U.S. National Counterintelligence and Security Center, told the AP: “Instead of dispatching spies to some parking garage in the U.S. to recruit a target, it’s more efficient to sit behind a computer in Shanghai and send out friend requests to 30,000 targets.”
In short, it’s an easy form of spycraft, because the potential source just accepts a connection from a pretty blonde and probably never thinks about it again until it’s time to congratulate the AI-generated person on their work anniversary.
For a while now people have been worrying about the threat of “deepfakes,” AI-generated personas that are indistinguishable, or almost indistinguishable, from real live humans. I think I may have caught an example of one in the wild:https://t.co/yvZbK8RoQt pic.twitter.com/4FaNqtivEY
— Raphael Satter @ RightsCon (@razhael) June 13, 2019