Ordinary Wi-Fi Signals Can Recognize People Without Cameras or Phones

Wi-Fi networks may be capable of recognizing people even when no camera is present and the person being observed is carrying no electronic device. Researchers at Germany’s Karlsruhe Institute of Technology developed a system that analyzes beamforming feedback information, or BFI—data that Wi-Fi devices continuously send to routers to help maintain a strong connection. According to reporting by New Atlas, the researchers used machine learning to interpret how a person’s body changes the radio waves moving through a room. Body shape, height and posture can produce subtle distortions in those signals. After being trained with recordings from known participants, the system was able to distinguish individuals by what the researchers described as a kind of radio signature. In a study involving 197 people, it correctly identified participants regardless of the direction from which they were viewed or the way they walked. Once trained, the system needed only a few seconds to tag a person. The method does not automatically reveal someone’s name. Like a fingerprint-recognition system, it first requires reference recordings connecting radio signatures to known individuals. The system can then determine which of those known participants produced a new recording. The privacy concern is that collecting the signal may not require access to a person’s own phone or laptop. New Atlas reported that BFI is transmitted without encryption and can potentially be recorded passively by a Wi-Fi adapter within range. Traffic between other devices and a router could provide the radio field altered by a person moving through the area. The researchers’ work suggests that changing one’s gait or approaching from another direction may not reliably defeat the system. It also raises the possibility of monitoring in places where Wi-Fi is already widespread, including homes, offices, airports, cafés and shops, without the visible hardware associated with conventional cameras. The team is calling for privacy protections to be incorporated into IEEE 802.11bf, an upcoming Wi-Fi standard covering wireless sensing. Its study, “BFId: Identity Inference Attacks Utilizing Beamforming Feedback Information,” was presented at the ACM Conference on Computer and Communications Security in Taipei. The researchers have also released their 197-person dataset for non-commercial research.