KAIST, NUS and SMU develop AI-based 'SweepLED' technology that detects hidden cameras with 94 percent accuracy using a low-cost LED device
"Just attach a 10,000 won ($7) LED device to your smartphone, and you can detect a hidden camera in five seconds."
KAIST announced Sunday that a research team led by Professor Han Jun of the School of Computing, in collaboration with the National University of Singapore and Singapore Management University, has developed SweepLED, a technology that detects hidden cameras by attaching an LED case to a smartphone.
Illegal spy cameras have grown increasingly small and sophisticated, making them easy to conceal in everyday spaces such as hotel rooms, short-term rentals and public restrooms. Conventional portable detectors require users to manually search for bright reflective spots from camera lenses — a method prone to false positives, as reflections from metal, glass or glossy plastic can be mistaken for a lens.
SweepLED takes a different approach. While the smartphone camera remains fixed, the direction of the LED light is varied, and an AI analyzes changes in light reflection across object surfaces — rather than simply looking for a "shiny dot."
On ordinary reflective objects, a reflection point shifts position or disappears depending on the direction of the light source. Camera lenses, by contrast, produce a distinctive reflection pattern when the lighting angle changes, owing to their internal structure — including the lens elements, aperture and image sensor.
The research team trained a deep-learning AI on the movement and shape of reflections over time as light was shone from multiple directions. Rather than having a person manually search for bright spots, the AI analyzes reflection patterns to determine whether a real camera lens is present.
The technology can also detect miniature cameras hidden inside everyday objects commonly found in accommodations or homes — including chargers, desk clocks, remote controls and decorative items.
In experiments on 30 objects found in real-world environments, the system achieved a detection accuracy of about 94 percent, with each object scanned in under five seconds.
Cost is another advantage. The core components of the LED case come to just $7, or about 10,000 won. The research team expects the technology to be developed into an affordable hidden-camera detector in the form of a smartphone accessory or a consumer security product.
"It can be used by individuals to check objects around them for hidden cameras in spaces where illegal filming is a concern, such as hotels, short-term rentals and changing rooms," Professor Han said. "Because it can be implemented as a small smartphone accessory and a dedicated app, ordinary users can use it conveniently whenever needed, without having to carry separate specialized detection equipment."
The findings, with KAIST doctoral student Yoon Jong-hyeok as lead author, were presented at ACM MobiSys 2026, an international conference on mobile computing, on June 20.
nbgkoo@heraldcorp.com