With the threat of AI and robots taking over the world, it may come as comforting news that computers aren’t immune to human trickery. Researchers at Fudan University in China, the Chinese University of Hong Kong, Indiana University, and Alibaba Inc. created a baseball hat that can fool facial recognition software. Humans: 1, robot overlords: 0.
Here's how it works. The inside of the hat projects infrared light onto strategic locations on the wearer’s face. The technique is called “adversarial learning,” and will trick a computer into thinking you’re Halle Berry. Well, maybe not Halle Berry, but maybe Moby... seriously. 70 percent of the time, the new technology was able to trick facial-recognition system FaceNet into identifying hat-wearers as celebrities, including the vegan musician. Before you get too excited, you should know that humans can’t see infrared light, so you won’t be able to catfish your next Bumble date.
In all seriousness, this development has important and potentially disastrous implications for security systems that rely on facial recognition. As the research paper notes, “Face-recognition techniques today are still far from secure and reliable when being applied to critical scenarios like authentication and surveillance.” Governments and companies would be wise to take note at the vulnerabilities of facial-recognition before becoming overly reliant on them. Meanwhile, the scary consequences could be a great source of inspiration for an episode of Black Mirror.