Biometrics via Wi-Fi: Signal disruptions allow identification and monitoring

Contents

Scientists at La Sapienza University in Rome have developed a technology that can secretly identify people simply by analyzing Wi-Fi signals – without any cameras or active involvement of the people concerned. This method, which poses great risks to privacy, makes use of the fact that every human body interferes with Wi-Fi signals in a unique way. The researchers have christened the technology, which initially sounded like dystopian science fiction, WhoFi, in reference to the international WLAN term WiFi.

The background to this is that every person leaves behind a kind of invisible Wi-Fi fingerprint. The researchers have developed a special model with artificial intelligence that evaluates these minimal signal changes in the form of channel status information.

Biometric characteristics such as body shape, size and movement are "extracted from channel state information (CSI) and processed by a modular deep neural network (DNN)" with a special encoder, the team writes in a paper that has not yet been reviewed by independent experts. The neural network is trained using a contrastive learning function to identify robust and generalizable biometric signatures. This approach helps the model to learn which data points are similar ("positive") and which are dissimilar ("negative").

In the context of WLAN devices, CSI also refers to information about the amplitude and phase of electromagnetic transmissions. According to the study, these measurements interact with the human body in a way that leads to person-specific distortions.

Recognition accuracy of over 95 percent

The scientists fed the network, which basically functions like a small computer brain, with information from the NTU-Fi data set. This is used as standard to detect the presence of people via Wi-Fi signals. According to the study, their own experiments indicate that the approach "achieves competitive results compared to state-of-the-art methods and confirms its effectiveness in identifying people via Wi-Fi signals." The system can therefore recognize with an accuracy of up to 95.5 percent whether someone is in a room and which person they are.

The technology works without visible surveillance and without the knowledge of the person concerned. This system could theoretically be used for personal identification wherever WLAN is available – whether in homes, offices or public buildings –.

Facial recognition: unreliable

The researchers describe the method as an alternative to conventional biometric recognition, which until now has mainly been based on camera images. In principle, this also works by measuring a person's unique biological or behavioral characteristics and comparing them with stored data. Automated biometric surveillance systems such as face and gait recognition use cameras and special software to identify people in a larger environment in real time without them having to interact with the system.

However, visual face or gait recognition systems are susceptible to external influences. Poor lighting conditions, obscured faces, changing viewing angles or low image quality can severely impair accuracy. Traditional biometric methods are therefore "unreliable", says the team. The new wireless biometrics are more robust as they use the unique signal distortions. These are caused not only by external features, but also by internal structures such as bones and organs. Distortions occur that could serve as an individual signature. In the next step, these biometric features are compared with known reference data to identify people as accurately as possible.

Raw data should be anonymous

While research in this field is still in its infancy, the development is already raising serious questions about data protection: Where is the line between technological progress and the right to privacy when even the invisible signals in and around people can be used for surveillance? Activists are already warning that automated facial recognition could lead to mass spying on citizens, which could take on entirely different dimensions in the future.

However, the scientists themselves also emphasize that the raw Wi-Fi data collected for personal recognition is anonymous by nature. This means that if it falls into the wrong hands, it is useless to attackers. Without the specially developed AI model and the associated system, it would not be possible to identify people.

Over the years, scientists have already discovered that Wi-Fi signals can be used for various sensor applications. For example, they are capable of seeing through walls, detecting falls, detecting the presence of people, and recognizing gestures, including sign language.

(nie)