That tape over your webcam may not be enough. Researchers at the Massachusetts Institute of Technology (MIT) have highlighted imaging privacy threats enabled by ambient light sensors, in a paper recently published in Science Advances. Device users concerned with security and privacy may be comforted by hardware solutions (shutters) and software permissions restricting webcam use. However, researchers have shown visual information can be gathered via one of the common ambient light sensors installed in many devices. These small sensors usually aren’t shuttered or disabled by users and are typically permission-free on a device level.
Ambient light sensors are categorized as low-risk by device makers and can often be accessed directly by software (or malware) without any permissions or privileges. Nevertheless, previous studies have shown such a rudimentary sensor can provide enough information to infer keystrokes on a virtual keyboard and steal a device PIN, about 80% of the time. The new research shows what an ambient light sensor can do when combined with an active light source component – namely the device' screen.
For their experiments, the MIT researchers used a Samsung Galaxy View 2. This rather old and large (17.3-inch) consumer tablet has its ambient light sensor next to the front-facing (selfie) camera, which is still a very common configuration.
In brief, the researchers found that the position and shape of a user's hand could be determined using their technology. Also, hand gestures could be revealed, as could the number of fingers used and gesture directions. Lastly, it was possible to detect other shapes, like the presence of a human face, for example.
The scientists explained that the ambient light sensor reads the light emitted by the screen shining on a person’s face and being partially blocked by the hand / screen interaction. A whole lot of complicated math, aided by AI and image processing technology, was used by the researchers to deliver their results.
Every device offers a variation in light sensor speed and measurement bit depth, screen brightness, and light sensor precision so some devices will be more vulnerable to this ambient light sensor spy technique than others (see figure above). In the source article figures you will see some of the image captures took several minutes on the tablet device used. However, ambient light sensor imaging snooping tech is demonstrably real and could be refined and optimized.
As ambient light sensors are quite useful, we don’t want devices to get rid of them due to security concerns. Instead, the MIT researchers propose the following adjustments to their implementation:
- Rethink ambient light sensor device permissions
- Reduce sensor speed
- Reposition the sensor so it doesn’t face the user
The above security-conscious changes could easily be implemented by device manufacturers without any obvious drawbacks, so we hope to see them adopted going forward.
