Your Body Distorts Wi-Fi, That’s How Hackers Can Track You

Scientists at La Sapienza University in Rome developed a new system to detect people through Wi-Fi signals, without the need  of cameras or phones.

The system, named WhoFi, captures how a person’s body distorts Wi-Fi signals and turns that pattern into a unique biometric signature.

“The core insight is that as a Wi-Fi signal propagates through an environment, its waveform is altered by the presence and physical characteristics of objects and people along its path,” the authors explain to The Register.

The recorded changes appear in Channel State Information (CSI), which contains detailed biometric information, like how bones and organs affect signal paths.

In their research paper, the team explains how they trained a deep neural network to recognize these body-specific distortions. Using a Transformer-based encoder, the WhoFi system achieved up to 95.5% accuracy on a public dataset.

The Register notes that that’s a major improvement over previous methods like 2020’s EyeFi, which reached around 75% accuracy.

In contrast to traditional video-based surveillance, which can struggle in poor lighting or with obstructed views, Wi-Fi-based identification is unaffected by darkness and can even see through walls. Researchers say this makes it more reliable and potentially more privacy-preserving, since it doesn’t capture images.

The authors  clarify that ‘re-identification’ technology does not expose personal  names, but it can confirm that one person has been spotted at different locations. The system enables tracking of individuals between Wi-Fi areas through their unique signal patterns, without carrying a device.

“The encouraging results achieved confirm the viability of Wi-Fi signals as a robust and privacy-preserving biometric modality, and position this study as a meaningful step forward in the development of signal-based Re-ID systems,” the authors say to The Register..

This new technology will likely raise new privacy and surveillance ethics concerns.