When people hear the word security, they usually imagine firewalls, antivirus software, access control systems or cameras.

I have been looking at the problem from a slightly different angle.

Can a computer detect that something is happening in a room without a camera, an infrared sensor or a radar?

In some cases, the answer is yes.

And sometimes, ordinary WiFi may be enough.

WiFi as a Sensor

Every obstacle changes how a radio signal propagates.

Walls, furniture, doors and people all affect reflection, attenuation, scattering and multipath behaviour. When a person moves through a room, the radio environment changes with them.

Modern WiFi chipsets can expose telemetry known as Channel State Information, or CSI. CSI describes how the wireless channel behaves across subcarriers, antennas and time. Research around WiFi sensing uses CSI to study how physical environments and human movement affect wireless channels.[2]

The important point is simple:

The system does not need to see a person. It only needs to observe that the radio environment has changed.

That makes WiFi sensing very different from a camera.

A camera records appearance.

WiFi sensing observes disturbance.

What I Mean by WiFi Fingerprint

The word fingerprint can be misleading.

I do not mean facial recognition, biometric identification or reliable recognition of a specific person.

By WiFi fingerprint, I mean the changing radio signature of a physical space.

A room has a certain radio behaviour when it is empty. It has another behaviour when a door is open, when furniture is moved, when a person walks through it or when multiple objects affect the signal path.

The goal is not to answer:

Who is this person?

The first and more realistic question is:

Has the space changed in a meaningful way?

That distinction matters.

Motion detection, occupancy change detection and anomaly detection are not the same as personal identification. They require different models, different levels of confidence and different privacy boundaries.

The Original Plan Was Simpler

When I first started thinking about this idea, I imagined it as a relatively small Linux service.

A supported WiFi adapter would provide useful telemetry. The service would read the signal data, compare it with a baseline and produce a security event when the environment changed.

That was the simple version.

The real version is harder.

It is not enough to read data from a WiFi adapter. A practical system would need to:

  • obtain CSI or comparable signal telemetry from supported hardware,
  • handle driver and chipset limitations,
  • filter noise,
  • compensate for environmental drift,
  • build a model of the physical space,
  • distinguish normal variation from meaningful movement,
  • avoid false positives,
  • and only then start thinking about more advanced pattern recognition.

This is not a weekend project.

It is an engineering and research problem.

Why CSI Matters

Traditional WiFi monitoring usually focuses on network behaviour:

  • signal strength,
  • connection quality,
  • roaming,
  • throughput,
  • packet loss,
  • authentication events.

CSI is different.

It can expose a more detailed view of how the wireless channel behaves. Instead of only asking whether the connection is strong or weak, CSI can help observe how the channel changes over time.

The Linux 802.11n CSI Tool is one example of research tooling that made CSI measurements available from specific Intel WiFi hardware on Linux, which is exactly the kind of hardware and driver dependency that matters for this type of project.[1]

That is why CSI has become important in research around device-free sensing, occupancy detection and human activity recognition. Research has explored cross-domain WiFi gesture recognition and device-free activity recognition, but those results also show why calibration, domain changes and environment independence are difficult problems rather than details.[3][4]

There is an important caveat.

Research feasibility is not the same as production reliability.

A lab result does not automatically become a stable security product. Real rooms change. People move furniture. Doors open. Access points are replaced. Firmware changes. Temperature, humidity and interference can alter the signal environment.

A useful system must survive those changes.

Why I Still Think It Is Worth the Effort

Despite the complexity, I think this direction is worth exploring.

Not because it replaces existing security systems.

It does not.

I see WiFi sensing as an additional signal layer.

A camera may show nothing because the movement is outside its field of view. A passive infrared sensor may miss slow movement or be placed in the wrong position. A door contact may only tell us whether a door opened, not whether the room behaviour changed afterward.

A WiFi fingerprint layer could provide another form of evidence:

Something in this space changed.

That signal may not be enough for a final conclusion. But it may be enough to trigger verification, raise confidence or correlate with other security events.

The future of physical security may not be one perfect sensor.

It may be several imperfect signals that validate each other.

Security Without Cameras

One of the most interesting parts of WiFi sensing is privacy.

Many organizations cannot or do not want to use cameras everywhere. Cameras create obvious privacy concerns. They capture faces, body movement, clothing, screens, documents and sometimes situations that should not be recorded at all.

WiFi sensing is different.

It does not store an image.

It does not know what a person looks like.

It does not recognize a face.

It observes changes in the radio environment.

That is a very different category of sensing.

But privacy-preserving does not mean privacy-free.

Even non-visual sensing needs governance. It needs purpose limitation, retention limits, access control, and clear rules about where it is used and why.

A system that detects movement without a camera can still be sensitive.

That is why this must be designed as a security architecture problem from the beginning, not as a gadget.

No Direct Line of Sight

Cameras require line of sight.

WiFi does not behave like a camera.

Radio signals reflect, scatter and propagate through environments in ways optical systems do not. That means a sensing layer based on WiFi may detect environmental changes in places where a camera has no useful view.

This can be useful in technical rooms, corridors, storage areas, restricted zones or spaces where visual monitoring is not appropriate.

Again, this does not make WiFi sensing magic.

It makes it complementary.

A camera tells us what it can see.

A door sensor tells us what opened.

A network sensor tells us what connected.

A WiFi fingerprint may tell us that the physical radio behaviour of the space changed.

Each signal has limits.

Together, they may provide better context.

A New Layer of Security Evidence

The key word for me is evidence.

I do not want to build a system that simply says:

Alert: movement detected.

That is too shallow.

A useful security layer should help answer better questions:

  • What changed?
  • When did it change?
  • Was the room expected to be empty?
  • Did the change correlate with a door event?
  • Did it correlate with network access?
  • Did it correlate with a system integrity change?
  • Is this a one-time disturbance or a repeated pattern?
  • What confidence do we have?
  • What evidence can we keep without violating privacy?

This is where WiFi sensing becomes more interesting than a simple motion detector.

It can become part of an evidence chain.

Not a single truth source.

A signal.

A correlation point.

A privacy-aware input into a wider security model.

Intelligent Buildings and Operational Context

The same idea is also relevant outside classic security.

An intelligent building may want to understand whether a room is occupied, whether a space is behaving normally or whether something unusual is happening after hours.

Possible future uses include:

  • detecting whether a room appears empty,
  • detecting movement in a restricted area,
  • observing occupancy changes,
  • identifying unusual after-hours activity,
  • correlating physical presence with access-control events,
  • supporting energy optimization without cameras,
  • and adding context to incident response.

The important word is supporting.

WiFi sensing should support decisions, not replace human judgement or stronger evidence where it is required.

Where This Fits Into FILIP:OPS / TCS

I am researching this as a possible future sensing layer for FILIP:OPS / TCS.

The goal is not to create another gadget.

The goal is to explore whether radio-environment telemetry can become part of a broader security and evidence architecture.

In that architecture, WiFi fingerprinting would not stand alone.

It would be one signal among others:

  • system integrity,
  • network events,
  • access-control data,
  • audit evidence,
  • operational baselines,
  • anomaly detection,
  • and human approval where needed.

If the system says that a protected space changed, that should not automatically mean an incident.

It should mean:

This is evidence worth correlating.

That is the kind of security layer I am interested in.

What Comes Next

There is still a lot of work ahead.

The hard parts are not only technical. They are also operational and architectural.

The system must be reliable enough to be useful, modest enough not to overclaim and privacy-aware enough not to become a surveillance problem.

That means working through:

  • supported hardware,
  • CSI access,
  • Linux driver behaviour,
  • calibration,
  • model stability,
  • false positives,
  • environmental drift,
  • deployment patterns,
  • evidence retention,
  • and governance.

The original idea was simple.

The real problem is much more complex.

But the deeper I go into it, the more I think this kind of unconventional telemetry will matter.

Modern security systems should not only observe packets, processes and log files.

They should also understand the physical context in which infrastructure lives.

And sometimes, the first sensor may already be in the room.

WiFi.

Sources

[1] Linux 802.11n CSI Tool: https://dhalperi.github.io/linux-80211n-csitool/

[2] A Survey on WiFi Sensing with Channel State Information: https://arxiv.org/abs/2207.04550

[3] Towards Environment Independent Device Free Human Activity Recognition: https://arxiv.org/abs/2006.10598

[4] Widar3.0: Zero-effort Cross-domain Gesture Recognition with Wi-Fi: https://ieeexplore.ieee.org/document/9007003