Swedish AI system detects concealed handguns using Wi-Fi signals with 72.5% precision

The Swedish Defense Research Agency has developed an open-source AI system called WiFireNet that can detect hidden handguns by analyzing ambient Wi-Fi signals. This presents a new non-invasive option that avoids intrusive cameras and bulky metal detectors in security screening processes. For buyers and security planners, this represents a passive, non-invasive detection capability.

Open-source WiFireNet uses older wireless standard for passive security screening

WiFireNet works by processing Channel State Information (CSI) from wireless signals. It detects objects based on how they alter ambient Wi-Fi wave patterns. The proof of concept system was developed for the older IEEE 802.11a standard. Presumably, future versions will use Wi-Fi 7 or Wi-Fi 8 standards.

In controlled tests, the system demonstrated an average precision of 72.5% in detecting handguns. The precision fell to 53.8% when the system was tested with stricter localization needs. The neural network struggles to perform accurately when encountering unfamiliar or untested indoor environments. This reflects the current limitations in complex and unstructured settings.

The system was released as completely open source by the research team. They published the source code, the trained models, and the custom training dataset. This enables other developers to examine the approach and expand on the system. The system classifies handguns versus other objects with deep learning algorithms.