"Wi-Fi can see you" makes a great headline and a poor explanation. The truth is more useful and more reassuring. Channel State Information is a measurement of how a room reshaped a radio signal, and that measurement reveals a specific, limited set of things. Knowing exactly where the line sits is what lets you use sensing without being creepy, and judge honestly when someone else's claim is overblown.
01What CSI genuinely reveals
CSI is sensitive, and it is fair to take that seriously. From a single Wi-Fi link read many times a second, you can detect:
- Presence — whether a person is in the area of the link.
- Motion — how much movement is happening, and roughly when it starts and stops.
- Coarse activity — broad patterns such as walking versus sitting still, because different movements reshape the channel in different ways. We cover this in activity recognition.
- Breathing rhythm — the slow chest motion of someone sitting still, in roughly the 0.1 to 0.5 Hz band.
That is genuinely powerful for presence and wellbeing uses, and it is genuinely personal information about the people in a space. None of it requires a camera, and all of it comes from the ordinary radio signal already passing through the room. The complete guide to Wi-Fi CSI explains the mechanism in detail.
02What CSI cannot reveal
Just as important is what stays out of reach, because this is where the scary headlines fall apart.
- Your face or an image. CSI carries no picture. There is no image sensor and no way to reconstruct a photograph of a person from a handful of subcarrier measurements.
- Your identity. A CSI stream does not know who you are. In a tightly controlled lab, research has distinguished a small set of known people by their movement signatures, but that is a narrow result in a fixed setup, not reliable recognition of a stranger walking in.
- Your screen, keystrokes or messages. CSI measures how the room changed a radio signal, not the contents of anything you type, display or send.
- Fine detail through a solid wall. Lower-band Wi-Fi passes through thin partitions with heavy loss, but it senses coarse motion at best, not a clear picture, as we discuss in the honest limits of single-antenna CSI.
03Why it is friendlier than a camera
For many jobs that have traditionally used a camera — knowing a room is occupied, triggering lights, noticing that someone has got up — CSI does the job while exposing far less. A camera captures a recognisable image of whoever is in frame, which can identify them, show what they are doing in detail, and leak catastrophically if the footage is breached. A CSI sensor captures none of that. There is no picture to steal, no face to match, and nothing that on its own points to a specific person.
This is why CSI and similar radio sensing are attractive for sensitive spaces such as bathrooms, bedrooms and elder-care settings, where a camera would be intrusive or unacceptable but presence and fall information is genuinely useful. The sensing is inherently lower-resolution about people, and for privacy that is a feature, not a flaw. It is the same reasoning behind choosing radar or CSI over a lens in the comparison in the 802.11bf sensing standard write-up.
04The legitimate concerns of ambient sensing
Friendlier than a camera is not the same as harmless. Ambient sensing raises real questions that deserve honest answers, not dismissal.
- It is invisible. A camera is at least a visible object you can see and cover. Sensing that rides on ordinary Wi-Fi gives no obvious sign it is running, so people may be sensed without realising it. That makes clear disclosure more important, not less.
- Presence patterns are personal. When someone is home, when they sleep, when a room is empty — a log of presence over weeks says a lot about a person's life, even with no image attached.
- Standardisation widens reach. As Wi-Fi sensing becomes a built-in network feature rather than a hobby experiment, far more spaces could be sensed by default, which raises the stakes of consent and data handling.
- Data can be combined. On its own CSI is coarse, but joined with other data — who owns the network, calendar entries, other sensors — it becomes more revealing than it looks in isolation.
The honest position is to hold both ideas at once: CSI exposes much less than a camera, and it is still data about people that can be misused if treated carelessly.
05Using it responsibly in your own space
The defensive, decent way to use Wi-Fi sensing is simple, and it starts from one rule: sense your own space, on your own network, with the knowledge of the people who share it.
- Stay on your own network and devices. Build sensing on Wi-Fi links you own and control. Do not try to sense other people's homes, networks, cards or signals — that is neither the point of this field nor acceptable.
- Tell the people who share the space. Anyone who lives or works where the sensor runs should know it exists and what it does. Invisible sensing of people who have not been told is the behaviour to avoid.
- Sense only where it is expected. A living room for occupancy lighting is reasonable. Aiming sensing at a guest, a neighbour or a shared wall to watch someone else is not.
- Keep the data small and local. Store only what the application needs, for as short a time as it needs, and process on the device or your own machine where you can. Less stored data is less that can leak.
- Prefer derived signals over raw logs. "Room occupied" or "fall detected" is far less sensitive than a continuous recording of everyone's movements. Keep the useful conclusion, discard the raw stream when you can.
Follow those and Wi-Fi sensing is a privacy-respecting tool for your own home: camera-free presence, gentle wellbeing monitoring, and automation that does not put a lens in every room. The technology is neutral; the responsibility is in how, where and on whom you point it — and the only decent answer is your own space, with consent.
06FAQ
Can Wi-Fi sensing identify who a person is?
Not from CSI in any ordinary sense. CSI records how the channel changed, not a face or a name, so it cannot recognise a person the way a camera and face matching can. Research can sometimes tell trained individuals apart in a fixed setup, but that is a controlled result, not reliable identification of a stranger.
Can Wi-Fi CSI read my screen or messages?
No. CSI measures how a radio signal was reshaped by the room between two devices. It carries no image, no text and no contents of what you type or display. It can sense that someone is present and moving, not what is on any screen.
Is Wi-Fi sensing more private than a camera?
Generally yes. There is no image to leak, no face to recognise and nothing to recognisably show a specific person, so a CSI sensor exposes far less than a camera pointed at the same room. It is still motion data about people, though, so it deserves care, consent and sensible storage.
How do I use Wi-Fi sensing responsibly?
Keep it to your own network and your own space, tell the people who share that space, and only sense where everyone present would reasonably expect it. Store as little as you need, process locally where you can, and never use sensing to watch other people, their homes or their devices without their knowledge and agreement.