Imagine this: you walk into an unfamiliar apartment and the first thing you do is check for any suspicious cameras. Nothing. There is an ordinary router sitting on the table, and you have not even connected to it. It seems there is nothing to worry about.
But researchers already need little more than that router to obtain far more information about a person than you might expect.
Wi-Fi stopped being merely a way to provide internet access a long time ago. Radio waves are constantly bouncing off walls, furniture, and people, and every movement slightly changes the signal pattern. If you learn how to detect these changes, you can determine whether a room is empty, whether someone is walking through it, or whether a person is standing still. In March 2026, researchers presented a system called WiAR that distinguished between walking, running, a stationary person, and an empty room with an accuracy of around 91%. And it did not need a single camera to do so.
But detecting that someone has walked past is not even the most surprising part. Modern experiments show that radio signals can potentially reveal who that person was as well.
Your Body Leaves a Trace in Wi-Fi
When we connect a laptop to a home network, it may seem as though the signal simply travels from the router to the device. In reality, the process is much more complicated. Radio waves spread throughout the room, reflect off objects and walls, travel along different paths, and only then reach the receiver. To them, a person is just another obstacle.
Our bodies partially absorb and reflect radio waves. Walk across a room, raise your arm, or sit down on a sofa, and the conditions in which the signal propagates change slightly. We do not notice anything, but to the system, these fluctuations become data.
One way to analyze them is through CSI, or Channel State Information. It shows what has happened to a Wi-Fi signal as it travels from the transmitter to the receiver.
In the case of WiAR, the researchers converted this data into images resembling spectrograms and then fed them into a neural network. The network learned to distinguish characteristic changes in the signal produced by different human actions. As a result, walking left one kind of radio-frequency “pattern,” running produced another, while an empty room looked different again.
Systems like these are of interest for smart homes. For example, they can determine whether someone has entered a room or fallen without requiring a video camera to be installed inside.
Radio signals also do not need light, which means the technology works at night.
Up to a certain point, this can even look like a more privacy-friendly alternative to conventional video surveillance. The system does not know what you are wearing or what expression is on your face. It simply detects changes in the radio signal.
The problem is that modern algorithms are gradually learning to extract far too much information from those changes.
The Router Recognized You
In 2025, researchers Julian Todt, Felix Morsbach, and Thorsten Strufe from the Karlsruhe Institute of Technology decided to examine another type of data generated by modern Wi-Fi networks. They called their system BFId.
At the center of the experiment was beamforming, a technology used since Wi-Fi 5. It helps a router transmit data more efficiently to a specific device: instead of simply broadcasting the signal in every direction, the system takes the state of the wireless channel into account and adjusts the transmission accordingly.
To make this possible, devices exchange special data known as Beamforming Feedback Information, or BFI. This information is necessary for the network to function properly and is transmitted wirelessly. The researchers discovered that these data retain a kind of imprint of the surrounding environment — including the presence of a nearby person.
Once again, the human body reflects and alters radio waves. But this time, the scientists went beyond simply detecting movement. Using intercepted data, they reconstructed representations of a person from different angles and then trained a model to distinguish one participant in the experiment from another.
The study involved 197 people. The system was able to identify them with accuracy close to 100%, and its performance remained consistent across different viewing angles and variations in gait. Once the model had been trained, the identification itself took only a few seconds.
And this is where the most unsettling detail of the experiment appears. A person does not have to use Wi-Fi themselves to become visible to a system like this. They do not need to connect to someone else’s network, enter a password, or even carry a smartphone. The method uses data exchanged between other Wi-Fi devices nearby. A person simply happens to be in the path of the radio waves, and their presence changes them.
The result is a rather unfamiliar form of surveillance. A camera at least has to see you. Here, it is enough simply to walk through a space where a wireless network is already operating.
Of course, BFId is still a research system tested under controlled conditions. This does not mean that every home router can automatically recognize everyone who walks past it today. But the experiment demonstrates just how much information can potentially be extracted from signals that already surround us all the time.
And Wi-Fi is far from the only way to do this. In 2026, another team of researchers decided to test an even more recognizable human characteristic.
It turned out that a system may not need a face, a voice, or a fingerprint to identify someone. Sometimes, it is enough to look at the way they walk.
When a Face Is No Longer Needed
A person can be recognized not only by their face, voice, or fingerprints. Each of us has another fairly stable characteristic as well — the way we walk.
We all place our feet differently, move our arms in different ways, hold our bodies differently, and have our own stride length and walking speed. These features are usually almost imperceptible, but to an algorithm they form a distinctive movement pattern. And a camera is not required to detect it.
In 2026, researchers from the University of Antwerp and the imec research center tested whether people could be identified using millimeter-wave Wi-Fi — radio waves at a higher frequency than those used in conventional home networks. Twenty people took part in the experiment, and the system analyzed changes in the signal as they walked. The result was a recognition accuracy of 91.2%.
The principle is similar to what we have already discussed, but the higher frequency makes it possible to capture more detail. When a person walks, different parts of the body move at different speeds and reflect radio waves differently. The legs produce one set of movements, the arms another, and the torso a third. The algorithm gathers these changes and looks for a combination characteristic of a particular person.
Millimeter-wave radar offers similar capabilities. In August 2026, a team from the University of Arizona described a system that uses radar data to reconstruct a kind of skeleton of a moving person and determine stride length and timing, walking speed, and other parameters. To test it, the researchers recruited 78 participants and conducted trials under four different conditions.
Instead of a photograph, the system produces a set of points and reflections. There is no face. No hair color, tattoos, or writing on a T-shirt either. But movement is clearly visible.
And this creates a new privacy paradox. Hiding your face from a camera is relatively easy. But stopping yourself from moving in your own characteristic way is much harder.
And radar can already detect much more than gait alone. In 2026 studies, it has been used to detect whether people are present, count how many are in a room, and recognize movements and poses. Experimental systems can distinguish whether a person is walking or remaining still, while some can also track their position in a room.
All of this sounds rather unsettling. But the most interesting thing is that many of these technologies were not originally created for surveillance at all.
The Camera Was Removed for Privacy. What Went Wrong?
Imagine the bedroom of an elderly person who requires constant monitoring because of the risk of falling. A camera can do the job, but few people would be comfortable with round-the-clock video recording inside their own home.
Radar offers a more discreet alternative. It can detect movement or a fall, but it does not record conventional video. It does not need light, so the sensor works at night, and no image of the room with faces and personal belongings is created at all.
That is why such systems are being actively studied for hospitals, nursing homes, and ordinary apartments. For example, in one 2026 study, millimeter-wave radar was used for continuous home monitoring: the system distinguished between a person being absent, remaining still, and walking, while also determining their position.
In another study, radar was trained to reconstruct the position of the human body without a conventional camera. Instead of video footage, the system receives a point cloud and uses it to determine the position of the joints. The developers explicitly consider this approach suitable for spaces where video surveillance is undesirable, such as medical facilities and smart homes.
It seems like an almost perfect compromise: the system knows enough to understand what is happening to a person, but not enough to actually look at them.
At least, that was the idea.
The better the algorithms become, the thinner that boundary gets. At first, a radio signal could tell whether anyone was in the room. Then it could determine what that person was doing. Then how they walked. And studies such as BFId have shown that under certain conditions, wireless signals can contain enough information to identify an individual.
This is where the idea of a “privacy-preserving sensor” becomes much more complicated.
A camera collects a great deal of obvious information: a face, clothing, the interior, the people nearby. A radio sensor does not see any of that. But if its data can be used to reconstruct movement, recognize actions, or distinguish one person from another, the absence of a photograph does not mean the absence of personal data.
Invisible Surveillance
For now, it is still too early to imagine turning every home router into a spying device. Many of the systems described above remain laboratory prototypes, work only under specific conditions, and require prior training. The results of an experiment involving 20 or 197 volunteers cannot automatically be applied to millions of people in a real city.
But the researchers from Karlsruhe point to another problem. BFId did not require a special radar to be built or a hidden camera to be installed. It used information that is generated during the normal operation of modern Wi-Fi networks. That is why the authors of the study are calling for protections against such forms of surveillance to be built into future Wi-Fi standards.
We are used to recognizing surveillance devices by sight. A camera looks like a camera. A microphone has to be hidden somewhere. A motion sensor can be spotted on a wall. With radio waves, that logic stops working. They are already all around us because they are needed for something completely different — connecting a laptop to the internet, opening an app on a television, or transferring data between devices.
Now it turns out that these same signals can also reveal information about the space they pass through.
Not long ago, if you wanted to know whether you were being watched, it was enough to look for a camera lens. In a world of wireless sensors, that method is gradually becoming outdated. A system may not see your face or create any conventional image at all, yet still detect that you entered a room, track your movement, and even try to determine exactly who you are.
It seems that the question “Where is the camera?” may soon be far from the most important one.
Share this with your friends!
10
Be the first to comment
Please log in to comment