For quite some time, digital surveillance seemed to depend on cameras, microphones, GPS, or mobile phones. Now, a German team has just demonstrated something that may cause unease: ordinary Wi‑Fi signals can be sufficient to recognize specific individuals even when they are not carrying devices.
Over the decades, wifi was understood as an infrastructure aimed exclusively at transporting data between devices. The researchers reveal that wireless waves contain traces enough to distinguish people; and not because cameras see them, but because their bodies alter the signal’s path in a characteristic way.
The conclusion is somewhat disturbing. The problem is no longer only what information we share on the internet, but the possibility that the physical environment itself begins to act as a silent sensor. Suddenly walls, air, and wireless traffic resemble less a passive network and more a system of environmental perception, turning an everyday infrastructure into a potential invisible biometric tool.
An environmental-perception system that turns an ordinary infrastructure into a potential invisible biometric tool.
The decisive detail was hidden inside the Wi‑Fi protocol itself
The work, presented at the ACM Conference on Computer and Communications Security, centers on a little-known technical component, called beamforming feedback information (BFI) or beamforming feedback information. This mechanism first appeared with Wi‑Fi 5 (802.11ac) to allow mobiles and computers to tell the router how to steer transmissions and improve coverage or stability.
That exchange happens continuously, and much of that information travels unencrypted. The research team identifies unique physical patterns within standard BFI traffic after studying how waves bounce off furniture, walls, and human bodies before returning to the access point. The idea vaguely resembles radar, though there are no military antennas or specialized equipment involved. The authors chose to use commercial routers and ordinary wifi cards to reconstruct wireless signatures associated with each participant through machine learning. The system was named BFId.
There is a relevant technical nuance. Earlier wireless identification works tended to rely on channel state information (CSI) or channel state information, a type of measurement much harder to obtain because it requires modified firmware and compatible hardware, such as the veteran Intel 5300 released in 2008. According to the paper itself, less than 6 percent of deployed devices supported CSI extraction in 2023.
Perhaps the most striking aspect is the method’s subtlety. No one needs to pose in front of a lens or unlock a phone. Simply existing within the space covered by the network suffices. Each torso, each shoulder, and each way of moving modifies the signal’s path in a way barely perceptible to us, but enormously descriptive for a trained artificial intelligence.
The experiment reaches nearly perfect figures with 197 participants
To test the system’s limits, the researchers analyzed 197 people in different scenarios. They describe this as the largest dataset used so far for this type of wireless identification. The artificial intelligence achieves recognition of individuals with a precision close to 99.5 percent even when they slightly change position or move around the room.
That doesn’t mean the experiment works like omnipotent magic. The model requires prior training, and the environment has a substantial influence. The furniture layout, the room’s geometry, or external interference all affect identification quality. Still, the performance achieved surprised even the authors themselves.
The technological leap isn’t that wifi perceives movement, but that it begins to recognize specific individuals.
There is another delicate aspect. The system does not merely detect “human presence,” something other wireless technologies have done for years. In this case, the network begins to discriminate specific identities. The conceptual difference is significant; detecting that someone is in a room is not the same as knowing exactly who it is. So the technological leap isn’t that wifi perceives movement, but that it begins to recognize specific individuals.
For quite a long time, biometric systems seemed tied to fingerprints, faces, or irises. However, this study suggests that the entire body can act as an involuntary radioelectric signature. The musculature, height, or even certain movement habits generate distinctive alterations within the wireless field.
The “noise” of wifi ends up functioning as a bodily fingerprint
The physics behind the phenomenon isn’t mysterious, though it is extremely complex. The electromagnetic waves change as they pass through different materials. Each person modifies the wifi signals following sufficiently characteristic combinations to betray their identity when an algorithm has abundant data. Because of that, the system can locate invisible regularities to a human observer. What to us seems like chaotic noise, for a neural network represents a collection of exploitable micro-variations. Machine learning stands out precisely for that: finding structures where there once seemed only disorder.
Curiously, BFId even outperforms CSI-based methods despite working with more compressed and, in theory, less detailed information. The authors argue that this compression acts as a kind of natural noise filter. Each BFI sample contains around 740 features, compared to the 212 typically used by CSI.
Add to that another decisive advantage. A single device in monitor mode can simultaneously capture information from multiple wifi clients and obtain multiple perspectives of a person within the space. Classic CSI-based approaches typically depended on a single angle per rogue node.
In parallel, the work partially debunks a widespread perception of digital privacy. Many people assume that turning off the phone or blocking geolocation dramatically reduces tech tracking. This experiment introduces a thorny caveat: even without active devices, the wireless environment can continue to describe our presence.
The researchers fear invisible uses in public spaces
The authors warn of potential applications related to covert surveillance, commercial tracking, or mass monitoring. The study envisions scenarios where wireless networks enable tracking people without visible cameras; precisely there lies much of the discomfort surrounding the project.
A shopping mall could tally visits without resorting to traditional facial recognition. An office could monitor internal movements with significant discretion. In authoritarian regimes, the technology could even be integrated into social-control systems that are difficult for the population to detect.
None of that seems imminent; although it is not entirely far-fetched either. Recent history shows that many tools originally designed to optimize infrastructures end up repurposed for different aims. It happened with mobile geolocation, with certain social platforms, and with numerous mass data analytics systems. Now, the question is not only what information we give to the internet, but how much the physical space we inhabit reveals about us.
The researchers warn of possible applications related to covert surveillance, commercial tracking, or mass monitoring.
The researchers themselves acknowledge important limits. The system does not “see” people as a camera would nor does it generate precise three-dimensional reconstructions. What it obtains is a statistical signature associated with specific signal perturbations. Several experts have pointed this out precisely to curb overly cinematic interpretations of the work.
Because of this, the team calls for incorporating specific protections into the IEEE 802.11bf standard, intended precisely for wifi sensing or detection via wifi. The goal is to prevent certain signals from being captured or reinterpreted easily by third parties.
Wi‑Fi stops being a simple network to become an ambient sensor
This project does not arise in isolation. For several years, different laboratories have explored ways to use wireless signals to detect breathing, domestic falls, human presence, or movements behind walls. The new generations of wifi are evolving toward systems capable of interpreting the physical environment in addition to transmitting data.
The difference now is that technology is crossing another boundary: from “there is someone here” to “we know who is here.” That nuance profoundly changes the debate. We are no longer talking only about home automation or technological convenience, but about identity and anonymity.
Perhaps the most singular aspect of the finding is precisely that. The Karlsruhe Institute of Technology study does not yet demonstrate inevitable total surveillance, but it does reveal that everyday wireless spaces contain far more information than we imagined. For quite a while we filled homes, stations, and offices with invisible waves to connect devices. We are only beginning to understand that these same signals could end up describing, following, and identifying the people who live within them.
