I don’t really worry about this kinda thing. A display has several GB/s of information displayed. Microphones don’t have nearly that much bandwidth. (This is why headphone cables don’t change much and Bluetooth audio is easy but video cables get updated every year with increasing display resolutions and wireless video is not as common).
For this to reliably work, they would effectively be compressing information out of the display signal.
In their paper they mentioned having to use size 175 font to distinguish between sets of letters typed on the screen. Which is cool, but pretty esoteric in my opinion.
A display has the capability to display several GB/s of information. But for the most part it's displaying a lot of static images and duplicate frames (unless you're watching video or playing games).
I don't think it's mathematically impossible to read smaller fonts if the same static window is on the screen for enough frames.
You can hear these sounds yourself using a carefully crafted pattern! This phenomenon was discovered on HN years ago and I made it into this prototype back then:
It's cool that even more normal images produce enough sound to detect and decode. The loudness varies a lot based on display though, lots of LCDs you can't even hear it at all even with the exaggerated screentunes pattern.
I'd actually prefer the mike in my phone, computer, etc., to not be that sensitive. I don't want it to be picking up background conversations, for example.
The only solution probably is to learn to hum or sing with the correct frequencies based on what it displayed to cancel the sounds of the display. This skill should be included in any IT security curriculum.
Also, please don't rewrite titles to editorialize. From the site guidelines: "Please use the original title, unless it is misleading or linkbait; don't editorialize." https://news.ycombinator.com/newsfaq.html
(Submitted title was "Your display contents may be traced by the sound of it using ordinary microphone")
10 comments
[ 3.3 ms ] story [ 31.5 ms ] threadFor this to reliably work, they would effectively be compressing information out of the display signal.
In their paper they mentioned having to use size 175 font to distinguish between sets of letters typed on the screen. Which is cool, but pretty esoteric in my opinion.
I don't think it's mathematically impossible to read smaller fonts if the same static window is on the screen for enough frames.
http://thume.ca/screentunes/
It's cool that even more normal images produce enough sound to detect and decode. The loudness varies a lot based on display though, lots of LCDs you can't even hear it at all even with the exaggerated screentunes pattern.
AND you can see around corners slightly
AND wifi can be used to find the layout of things in the room
So overall, you can analyze a lot more than we previously were able to.
Just wait until your very thoughts are deduced from 3 miles away
Also, please don't rewrite titles to editorialize. From the site guidelines: "Please use the original title, unless it is misleading or linkbait; don't editorialize." https://news.ycombinator.com/newsfaq.html
(Submitted title was "Your display contents may be traced by the sound of it using ordinary microphone")