I Always Feel Like Somebody’s Watching Me
The good news is that tiny video cameras are now widely accessible, inexpensive, and completely unintrusive. The bad news is that tiny video cameras are now widely accessible, inexpensive, and completely unintrusive. There are two sides to every coin, and those same cameras that keep your home and business secure, or just let you check up on your pets while you are away, can also be used to covertly spy on you in sensitive locations such as hotel rooms or bathrooms.
While there are ways to detect nearby cameras, doing so typically requires both expertise and specialized equipment — and may also rely on assumptions, such as wireless streaming of camera data over WiFi. These types of solutions do not have the average person that just wants to make sure they are not being spied on in their hotel room in mind.
A team of researchers at National University of Singapore and Yonsei University have leveraged advances in smartphone technology to develop a hidden camera detector that is easy enough for Grandma to use. Well, easy enough for the sort of grandma that has one of the latest smartphone models, anyway.
The method, called LAPD (Laser Assisted Photography Detection), makes use of the time-of-flight (ToF) sensor present on many newer smartphones, including Apple iPhones and Samsung Galaxy devices.
These ToF sensors normally emit a beam of laser light to determine how far away objects are. LAPD repurposes the ToF sensor to observe reflections of the laser light, to search out unique, unusually intense, reflections that are characteristic of the lenses of hidden cameras.
A phone app was designed to allow users to interact with LAPD. After choosing a suspicious area to scan, the app guides the user through the process of performing a scan. The processing pipeline then uses a custom algorithm to detect high-intensity reflections, which are filtered by shape and distance to identify suspicious areas.
A deep learning filter, consisting of a convolutional neural network implemented in TensorFlow Lite, is then used to remove false positives. Any remaining areas of concern are marked on the screen such that the user can manually examine that specific location.
The team tested LAPD out on several recent Samsung devices with onboard ToF sensors. Overall, the technique was able to detect 88.9 percent of hidden cameras. Using the naked eye alone, only 46 percent of hidden cameras were able to be detected. Thanks to the deep learning filter, LAPD will not be sending you on too many wild goose chases either — the false positive rate was found to be 16.67 percent.
LAPD offers an easy to use solution to the problem of detecting hidden cameras, but is not without its drawbacks. While ToF sensors are convenient to use because they are being incorporated into many smartphones, they are far from being in all devices at this time. Further, due to peculiarities in how ToF sensors work, only one object can be scanned at a time, which can make the process of scanning an entire room fairly lengthy.
These issues notwithstanding, LAPD is a big step in the right direction for helping the average person to ensure their privacy in a world where privacy is ever more difficult to come by.