In a paper, researchers have highlighted how Wi-Fi signals and AI can be used to track humans even behind a wall.
Who-Fi utilises a transformer-based neural network to create human signatures
Photo Credit: Unsplash/Dreamlike Street
Who-Fi is a cutting-edge technology that leverages artificial intelligence (AI) to identify and track individuals without needing any visual input. It is an experimental technology which remains to be thoroughly tested in the real world. However, based on the research paper that mentions and documents its proof of concept, it can be used to turn any ordinary Wi-Fi signal into a biometric scanner that can not only track the movement and active position of an individual but also identify their unique biometric signature.
According to a paper published in the online preprint journal arXiv, regular 2.4GHz Wi-Fi signals can be used to identify and track individuals, and play an important role in both identity authentication and surveillance. The technology also raises fresh concerns about digital privacy and security.
Breaking the technology down, the Who-Fi system uses a combination of a Wi-Fi signal and a transformer-based neural network (also known as a large language model). This LLM analyses and understands something known as “channel state information” (CSI). It monitors the changes in Wi-Fi signal strength and phase as they bounce around a room and reflect off an individual's body. This can be understood as the signals transmitted by radar and sonar systems.
So, whenever a human is near a Wi-Fi signal, the distortion in the signal's natural path creates a unique pattern. This pattern is said to be as accurate as other biometric signatures of humans, such as fingerprints, facial patterns, and the structure of the retina. The Who-Fi system can recognise this signature and attribute it to individuals.
Once it has been trained on these signatures, the system can not only track the individual's movement, but also identify them even when they re-enter the network zone after a prolonged period. It can also capture body movement data and recognise sign language. The main advantage of the system is that it functions without any visual or auditory sensors, such as cameras and microphones.
As per the study, the entire Who-Fi system requires a single-antenna transmitter and a three-antenna receiver, making deployment inexpensive. Coming to the efficiency of the system, the researchers found that despite the target being behind a wall, walking at a normal speed, Who-Fi achieved 95.5 percent precision.
The accuracy is said not to change even if the individual changes clothes or wears a backpack. Notably, a single system is capable of identifying and tracking up to nine individuals simultaneously.
Who-Fi also has high evasion, meaning it is very difficult to be detected by technologies that spot surveillance technologies. This is possible as the system does not use any special hardware, and there is no special emission pattern, infrared, radar, or visible spectrum light that can be detected. Additionally, Who-Fi conducts passive radio frequency (RF) sensing, making it easy to stay hidden.
Catch the latest from the Consumer Electronics Show on Gadgets 360, at our CES 2026 hub.