Picture the scene: you're in the middle of a regular video-conference session with your boss and colleagues, no-one is paying attention to you, and it's become tedious. So you bring up Notebookcheck News on the machine you're using to have something worth looking at while the meeting washes over you. However, the nightmare scenario of somehow being called out on your boredom-smashing behavior is always at the back of your mind.
This possibility may now become a reality, thanks to new, partially Defense Advanced Research Projects Agency- (DARPA) funded research into side-channel exploits. It suggests that cyber-spying during video calls is perfectly doable. This project has resulted in Synesthesia, a method that depends on picking up audio through the microphone of target devices for a certain portion of coil whine.
This coil whine is the high-pitched audio that pertains specifically to the power used by an LCD screen to display pixels. The researchers behind Synesthesia found that it could be isolated and converted into a visual form not dissimilar from a raster plot.
This was an impressive achievement. However, the graphical data was still complex and difficult to relate to specific screen contents using conventional means. This is where machine learning came in. The researchers reported that their neural networks were capable of processing the data with greatly improved success. Furthermore, it was able to match it to the typical display contents associated with the websites ranked 1 to 10 on alexa.com with an accuracy of nearly 97%.
Therefore, managers worried about what staff are really doing with their time may soon be able to monitor the activity - all of it - on a corporate video call. In addition, the existence of Synesthesia is yet another worrying sign of how privacy can leak out of any communications-related peripheral.