Surveillance cameras can keep us safe or invade our privacy, and improvements in miniaturisation mean they can now be hidden in almost any object. Researchers have developed a smartphone app that is claimed to detect such hidden cameras with 88.9 per cent accuracy.
Commercial devices to spot hidden cameras are available, but carrying a specific device for this purpose is inconvenient. at the National University of Singapore and his colleagues hope to provide the same function via a standard smartphone.
The system uses a time-of-flight sensor, which emits a beam of infrared light and measures how long it takes to be reflected from an object. The sensors have become increasingly common in smartphones and are used to improve the quality of photographs.
Advertisement
The researchers created a smartphone app that uses information from these sensors and machine learning techniques to spot the unique reflections from hidden cameras.
In tests, they found that if the object is scanned at too short a distance, the sensor will oversaturate with returned light, while if the phone is too far away, the reflections are too weak. To get around this, they added augmented reality features to allow the user to choose an object to check for cameras and get feedback on the optimum distance as they scan it from all angles.
The app then observes any spots on the object where the incoming light is brighter, which indicates a reflective surface and potentially a camera lens. It uses machine learning techniques to separate innocuous results from suspicious ones.
The researchers recruited 379 people and showed them videos of 30 different objects recorded from less than a metre away. They had to identify if they saw any hidden cameras and mark the location of them. With the naked eye, the volunteers only succeeded at spotting the cameras 46 per cent of the time, whereas the AI tool had a success rate of 88.9 per cent.
鈥淯sers can trade off between catching more hidden cameras but dealing with more false positives and vice versa just by adjusting a slider,鈥� says Sami. 鈥淟onger term, there are very promising avenues [for improvements]. The most important one is to also use the smartphone鈥檚 flashlight to induce reflections from hidden cameras, and detect them with the standard smartphone camera.鈥�
Raj Bhan at spy camera retailer Lorraine Electronics says he does see customers looking for detection equipment after having read news stories about hidden cameras, but he is sceptical that a smartphone could rival a commercial device designed for this purpose. He believes that false positives will be hard to avoid.
Proceedings of the 19th ACM Conference on Embedded Networked Sensor Systems
