Researchers have developed a system that can identify you using nothing more than the way your body disturbs a Wi-Fi signal — no device, no image, no consent required.

Their system, aptly named WhoFi, needs only the signal disruption your body causes when you walk through a room with Wi-Fi. And it works astonishingly well.

“The core insight is that as a Wi-Fi signal propagates through an environment, its waveform is altered by the presence and physical characteristics of objects and people along its path,” the researchers wrote in their study, published in arXiv. “These alterations … contain rich biometric information.”

Our privacy is dwindling away

Our sense of privacy was already eroded. From facial recognition in public spaces to apps tracking our every move, the tools that watch us have become more advanced — and harder to detect. But while most security systems use your phone, face, or password to identify you. Wi-Fi sensing uses how you move through the air.

Every time a Wi-Fi signal travels through a room, it bounces off walls, furniture, and, yes, human bodies. The result is a cascade of tiny alterations in signal amplitude and phase, known as Channel State Information (CSI). This data is usually used to troubleshoot wireless connections. But in the right hands, it becomes a personal signature.

The WhoFi system takes these invisible perturbations and feeds them into a deep neural network, trained to recognize and remember how each body uniquely disturbs a signal. It doesn’t need to interact with a person in any way. It just needs them to walk by.

Using a public dataset called NTU-Fi, which records how 14 volunteers altered Wi-Fi signals while walking in different clothing configurations — with and without backpacks, jackets, and coats — the researchers found that their system could correctly identify individuals with up to 95.5% accuracy.

“Privacy-Preserving” or Privacy-Threatening?

At first glance, WhoFi’s creators present their invention as a more privacy-conscious alternative to camera-based systems.

“By leveraging non-visual biometric features embedded in Wi-Fi CSI,” they write, “this study offers a privacy-preserving and robust approach for Wi-Fi-based Re-ID.”

They point out that Wi-Fi signals don’t capture images, record conversations, or see what you’re wearing. They simply capture the way your body subtly bends radio waves. And unlike cameras, Wi-Fi sensing works in darkness, through walls, and in low-visibility conditions.

But this same strength raises immediate red flags.

The potential for misuse is as vast as the modern wireless landscape. Wi-Fi networks are nearly everywhere and as more devices adopt Wi-Fi 6, 7, and 8 — which support finer-grained CSI measurements — this kind of tracking becomes even easier.

Unlike a camera on the wall or a security guard at the door, Wi-Fi surveillance can be invisible. You may never know it’s there and it could track you all the same.

Contactless Surveillance

To understand what makes WhoFi groundbreaking, it helps to look at what came before. In 2020, the same team introduced EyeFi, an earlier attempt at Wi-Fi-based identification. That system topped out at about 75% accuracy. WhoFi dramatically improves on that — pushing above 95% with Transformer models and customized training techniques like “in-batch negative loss,” which helps distinguish between many individuals simultaneously.

The team also tested older methods like LSTM (Long Short-Term Memory) and Bi-LSTM models, but the Transformer consistently outperformed them, especially when analyzing longer sequences of signal data. More training didn’t just make the model smarter — it made it more discerning.

“The Transformer-based model delivers the best overall performance,” they write, citing its self-attention mechanism, which helps it pick up on long-range patterns in how a person’s movement alters a signal over time.

Crucially, their model didn’t require any pre-tagged visual images or wearable sensors. It simply learned from the unique way each participant’s body disrupted Wi-Fi.

So What’s Next?

For now, the researchers emphasize that their work is academic. There are no commercial plans, no government rollouts. Yet.

But Wi-Fi sensing is a growing field. It’s already being explored for applications in fall detection, gesture recognition, presence sensing, and even measuring heart rate. The IEEE has ratified standards for “Wi-Fi Sensing” under the 802.11bf specification, meaning the industry is preparing for broader adoption.

That means the leap from lab to living room (or airport, or school) could be short.

“The encouraging results achieved confirm the viability of Wi-Fi signals as a robust and privacy-preserving biometric modality,” the authors conclude. “And position this study as a meaningful step forward in the development of signal-based Re-ID systems.”

But it also raises a simple question: If Wi-Fi can recognize you without your knowledge, is anonymity still possible?