Ordinary WiFi has learned to recognize a person with almost 100% accuracy

Scientists at the Karlsruher Institut für Technologie used these signals to reconstruct how radio waves propagate through space and reflect off people. The result is an image similar to a photograph, only made not with light, but with radio waves. Then machine learning compares such “radio portraits” and identifies a specific person.

The experiment involved 197 people. After training, the model identified them with almost 100% accuracy. It did not matter from which angle the person was observed or how the person walked. What matters is something else: the person does not need to carry a phone or smart watch. It is enough that other people’s WiFi devices are operating nearby.

The researchers emphasize the risk. Routers are already installed at home, in cafés, offices, and public places. A network that looks harmless could, in theory, become an invisible surveillance system. Unlike a camera, it does not draw attention.

The authors call for privacy protection to be built into the future WiFi standard IEEE 802.11bf before this capability becomes widespread.

Sources: Karlsruher Institut für Technologie (KIT). "Ordinary WiFi can now identify you with near-perfect accuracy." ScienceDaily (August 12, 2026).

Todt J., Morsbach F., Strufe T. "BFId: Identity Inference Attacks Utilizing Beamforming Feedback Information." ACM Conference on Computer and Communications Security (2026). DOI: 10.1145/3719027.3765062.