Wi-Fi sensing through walls: what is physically real, and what is not

2026-10-11 · 2 min read

"Wi-Fi can see through walls" is the headline that made Wi-Fi sensing famous. It is also the claim most often misunderstood. Here is the separation between physics and marketing.

What actually happens

1. A Wi-Fi signal leaves the router and reaches your client plus dozens of reflected paths;
2. Some paths pass through walls or doors and bounce off objects (and people) in the next room;
3. When someone moves behind a wall, those paths change slightly → the measured CSI changes;
4. A receiver-side algorithm notices the change — that is the "seeing".

What you get is a motion signal, not an image. You cannot reconstruct a body, a face, or a text message. With many antennas and a controlled setup, research can reconstruct coarse pose — but that is research, and largely a preprint today ^[6].

The motion → channel-change mechanism itself was demonstrated rigorously as early as 2013 (a custom MIMO array detecting motion through a wall) ^[3].

Attenuation reality check

Material attenuation in this frequency range has been measured and modelled in peer-reviewed work ^[4].

Also: the transmitter side matters. If the person is between the router and the client, the signal is strong; if they are two walls away, you are relying on weak reflected paths.

What research actually shows

• Research such as "DensePose from WiFi" ^[6] reconstructs coarse human pose from CSI — with multiple antennas, controlled setups, and heavy models. It is not a consumer feature.
• Papers on "through-wall" sensing typically report results in single-family homes with drywall, and accuracy drops sharply with concrete or metal.

The privacy side (and why you should write about it)

Wi-Fi sensing is easy to deploy covertly: a router or a laptop can sense a room without a camera. That is why research on RF privacy exists — for example the "CSI-MURDER" project, which studies how to jam or mask CSI-based sensing ^[5] — and why standards work (IEEE 802.11bf for WLAN Sensing ^[2]) treats privacy as a design question (survey context: ^[1]).

If you publish Wi-Fi sensing content, include the ethical frame:

• only sense spaces you own or where people consented;
• state clearly that behaviour can be inferred even without images;
• mention mitigation options in shared spaces.

Practical takeaways

1. Motion detection through a single drywall: realistic, and our capture shows it clearly in the amplitude heatmap;
2. Presence of a still person through a wall: unreliable (needs respiration micro-motion);
3. Imaging/identity: not a consumer capability — treat such claims as marketing.

Related: Wi-Fi CSI radar · explained · dataset