Researchers Develop AI to Detect Hidden Pistols Using Wi-Fi Signals
Swedish researchers have created an AI system capable of identifying hidden pistols by analyzing Wi-Fi signal disruptions. This non-invasive method could serve as an alternative to traditional security screening.
Researchers at Sweden's Defence Research Agency (FOI) have developed an artificial intelligence system that can detect hidden handguns by analyzing how Wi-Fi signals are altered by their presence. This non-invasive technology, named WiFireNet, offers a potential alternative to conventional methods like visual cameras and metal detectors.
The system operates by analyzing Channel State Information (CSI), which captures how objects disturb Wi-Fi signals through reflection and diffraction based on their shape, orientation, and material. To build WiFireNet, researchers first compiled a dataset called WiFire Dataset, recording synchronized RGB video and Wi-Fi CSI data from individuals carrying various items, including firearms. A deep learning model was trained on this data to map the visual characteristics of objects to their corresponding Wi-Fi signal perturbations.
Once trained, WiFireNet can identify hidden pistols using only Wi-Fi CSI, although accuracy varies. In controlled indoor experiments tracking up to four moving individuals in confined spaces, the average accuracy for pistol detection was 72.5 percent. More stringent localization requirements, aiming to pinpoint the weapon's exact position on a person, reduced accuracy to 53.8 percent. The system's effectiveness also decreased when encountering unfamiliar indoor layouts.
The project is currently in an experimental phase, with plans to enhance performance by utilizing modern Wi-Fi 7 and Wi-Fi 8 standards for higher-resolution CSI data. The research team has open-sourced the code, models, and dataset to encourage further research. While Wi-Fi sensing technology is gaining attention for its potential in surveillance due to low deployment complexity, concerns exist regarding privacy implications and the lack of regulatory frameworks for widespread use.