Researchers at Germany's Karlsruhe Institute of Technology (KIT) have shown that ordinary Wi-Fi traffic can identify individuals with almost 100% accuracy.
The system, called BFId, needs no camera, no special hardware, and no phone carried by the person being watched.
Wi-Fi is a form of electromagnetic radiation on the same spectrum as visible light, but at a much lower frequency.
That lower frequency lets it pass through walls, where light cannot, because it interacts less with the electrons in solid materials.
Bouncing from the ceilings
Even so, a portion of every Wi-Fi signal bounces off walls, ceilings, floors, and anything else in a room.
Devices pick up several reflected copies of the same signal, arriving fractions of a second apart, and together these form a distinctive interference pattern.
Because the human body is mostly water, and water absorbs Wi-Fi, a person walking into a room subtly reshapes that pattern.
The exact distortion depends on a person's size, shape, and movement, in effect turning them into a walking radio fingerprint.
More data
Wi-Fi routers already measure these patterns constantly to keep connections running smoothly, generating far more data than most users realise.
The KIT team fed this data into a machine learning model and trained it to recognise individuals within seconds.
In tests on 197 participants, the largest such study to date, the system correctly identified people regardless of the angle they were viewed from or how they walked.
There are benign uses for this, such as detecting a fall in a room without installing a camera.
There is an obvious downside too: the same technique could identify someone simply walking past a cafe window.
Policy safeguards
The researchers are now pushing to have privacy safeguards built into IEEE 802.11bf, the next Wi-Fi standard that formally introduces this kind of sensing.
Their argument is that this may be one of the last moments to build protection in before the capability becomes routine.
The bigger question the study raises is not whether this technology works, but whether anyone gets a say in switching it off.
