Spy beyond walls using WiFi signals

Monitoring of WiFi signals, together with processing techniques can be used to detect human activity inside buildings. These techniques represent alternatives to classic RADAR systems, and allow you to spy on the activity of a person (walking, running, jumping) through the analysis of WiFi signals. As a result, a simple commercial WiFi router can be used as a sensor to spy on and obtain information on moving objects or people.

Spy on activities using channel state information

Human motion data can be extracted from Channel state information (CSI). This information describes how a signal propagates from the transmitter to the receiver and represents the combined effect of e.g. scattering, fading, power decay with distance, etc.

In a WiFi transmission the channel parameters (amplitude and phase) are continuously estimated for all sub-carriers of the transmitted signal. This information is calculated for each received packet, and is collected by the CSI in a large complex matrix that describes the frequency response of the channel. The information depends on the surrounding environment, so if at each path it is possible to estimate the amplitude and phase of the signal, then it is possible to know the activity within a certain place.

For example, you can see how amplitude and phase change in two different scenarios. In fact, in the following figure you can see the amplitude and phase changes for an empty room (left) and a room with a moving person (right diagrams). The presence of a person induces changes in channel parameters that can be exploited by Human Activity Recognition (HAR) algorithms.

Spy on activities thanks to the doppler effect of WiFi

A person in motion causes scattering phenomena that depend on the speed of movement of the individual. The movements cause variations in the time it takes for the signal to reach the receiver through each of the propagation paths. This is reflected in a phase shift in the received signal. That is, the movements related to the activity of a human being cause phase variations , since each part of the body acts as a sink that moves at a specific speed.

In particular, this technique uses the Doppler effect of the WiFi signal and it is possible to observe the differences of the Doppler traces obtained when a person performs activities inside a room, compared to a trace taken in an empty room. The Doppler trace relating to the empty room has low power levels which reveal precisely the lack of movement in the environment.

Recognition of activities with neural networks

The Human Activity Recognition algorithm contains two basic steps and uses the amplitude and phase information, combining these data for a certain number N of available receiving antennas.

• First, Doppler traces are computed from the collected data of all receiving antennas to obtain activity estimates through an algorithm based on neural networks . In particular, machine learning-based prediction is applied to the data stream from each of the available antennas.
• As a result of the previous step, N independent predictions are obtained, one for each antenna, which are combined, in a second step, through a decision merging method that leads to the final estimate of the activity.

The recognition accuracy is over 95%, reaching almost 100% when the environment and location remain the same as the training data , regardless of the measurement day and the person doing the activity. The following figure shows the results on a confusion matrix for a test with the parameters: environment, day and person changing from the training.

What we have seen so far is the result of an all-Italian research conducted in collaboration between the University of Palermo and the University of Padua. The study is of particular interest, as it is possible to recognize the activities of a person inside a room with a simple commercial WiFi router.

Seeing through walls with WiFi: the MIT device

A more complex system using a multiple-input and multiple-output (MIMO) radio with an antenna array was developed at the Massachusetts Institute of Technology (MIT) with surprising results.

Finally, the use of these techniques poses privacy problems, since environmental activities are clearly identifiable through a common WiFi signal. The scientific world proposes solutions that present some research directions to mitigate the risks. For example, a solution could be to intervene on the transmitter with signal processing techniques in order to make a variable CSI appear in reception even when the environmental conditions are static.

The article Spying Beyond Walls Using WiFi Signals was written on: Tech CuE | Close-up Engineering .