Researchers develop AI to detect handguns and identify individuals via Wi-Fi signals

New research highlights the dual potential of Wi-Fi signals for both security detection and privacy intrusion. Researchers at the Swedish Defence Research Agency (FOI) have developed an experimental AI system called WiFireNet. By analyzing Channel State Information (CSI)—the way objects distort wireless signals—the system achieved a 72.5% average accuracy in identifying concealed handguns in controlled indoor environments.

Separately, a German research team presented findings at the ACM CCS conference regarding identity inference attacks. Their study, titled ‘BFId’, demonstrates that attackers can use unencrypted Beamforming Feedback Information (BFI) to create a ‘radio signature’ of individuals. Because human bodies cause unique geometric distortions to radio waves based on height, build, and gait, machine learning models can identify individuals with high precision without the need for cameras or password cracking.

The researchers emphasize that these vulnerabilities exist because BFI data is currently transmitted in plaintext. They are calling for mandatory encryption and privacy protections in the upcoming IEEE 802.11bf (Wi-Fi Sensing) standard to prevent these invisible tracking capabilities from being exploited.