Thanks to a few cheap Wi-Fi routers and artificial intelligence, it is possible to spy on someone through the walls of their house, or to detect an intruder… We explain how it works.
Researchers at Carnegie Mellon University have developed a monitoring system based on Wi-Fi signals. This device is able to determine position and movements human bodies in real time, even through building walls. As the 3D images shared by the researchers show, it is possible to observe the movements of several people without any problem.
How do I observe someone using Wi-Fi?
Concretely, the technology analyzes the way in which bodies interfere with Wi-Fi signals. Like a piece of furniture or a wall, a person disturbs the propagation of the signal emitted by a router. Depending on its size, shape and thickness, a body affects the signal differently, which amounts to a kind of fingerprint.
Disturbances caused by bodies can be intercepted through the channel state information (CSI). It is a set of data that indicates how a signal travels from the transmitter to the receiver. This information makes it possible to adapt the transmissions to the way in which the signal propagates. During the tests, several affordable routers, the TP-Link Archer A7 AC1750, were placed at each end of the room. Once the equipment was installed, people moved through the space covered by the Wi-Fi network.
Read also : A drone can spy on you through walls because of a Wi-Fi flaw
An artificial intelligence that learns
To analyze the interference generated by human bodies, researchers rely on artificial intelligence. A deep artificial neural network has been developed. It evolves and learns in the same way as the human brain. At the end of its apprenticeship, the AI was able to understand and detect the presence of a body in the middle of Wi-Fi signals.
As part of the development of the algorithms, Carnegie Mellon experts used DensePose, an AI created with the help of Meta. It was designed for map the body of a human being in 3D from a photo or video. This algorithm open-source helped artificial intelligence move from a simplistic 2D representation to an accurate 3D illustration of people in the field of the Wi-Fi network. To facilitate data analysis, researchers broke down the human body into 24 segments different. The approach made it possible to speed up the processing by artificial intelligence.
The researchers point out that their solution is not yet fully developed. There is still misfires from time to time, especially when an individual makes unusual movements, which are more difficult for artificial intelligence to grasp. Malfunctions also appear if the room accommodates too many people.
Efficient, cheaper and less intrusive
The innovation aims to be more reliable than a traditional surveillance camera. Unlike a camera, the combination of Wi-Fi and AI can spot a person who hides behind a piece of furniture or a partition. Moreover, the detection does not depend on the ambient luminosity. It is possible to detect a body in complete darkness. This technology is also less expensive than tools such as radars or Lidar sensors, the American researchers point out:
“Radar and lidar technologies require specialized, expensive and power-demanding hardware.”
Unsurprisingly, the researchers do not intend their creation to spy on individuals, comfortably seated on their sofa. Rather, this system is designed with a view to detect possible intrusions, like a burglar in a house for example, without compromising the privacy of the occupants. For engineers, their technology ” paves the way for low-cost, widely accessible and privacy-preserving algorithms”:
“Most households in developed countries already have Wi-Fi at home, and this technology can be put in place to monitor the well-being of older people or simply identify suspicious behavior at home.”
This isn’t the first time Wi-Fi has been used as a remote monitoring device. In 2016, researchers from theNorthwestern Polytechnic University in China demonstrated similar technology. This device also made it possible to monitor the presence of one or more people with Wi-Fi, through the walls of a house. Chinese innovation went even further by guessing the identity of an individual using the properties of his body. The system could indeed identify a member of a family of six with an accuracy of over 85%.
On the contrary, the technology developed by Carnegie Mellon University is not able to identify a person who walks in the area covered by the Wi-Fi. This is why their solution combining Wi-Fi and AI is “respectful of privacy”note the engineers.
Cornell University
