Researchers train AI to detect concealed handguns via Wi-Fi signals

Researchers at Sweden's FOI defense agency built WiFireNet, an AI system that identifies concealed handguns from how objects distort ambient Wi-Fi channel state information. In controlled trials it reached 72.5% average detection precision, dropping to 53.8% when pinpointing the weapon's exact location. The code, models and WiFire Dataset were released as open source.

• Average handgun-detection precision reached 72.5% with up to four moving people
• Precision fell to 53.8% when localizing the weapon's exact position
• Prototype uses the 1999 IEEE 802.11a standard; Wi-Fi 7 and 8 planned next
• Code, models and the WiFire Dataset were released openly