I think the photoacoustic effect is at play here. Discovered by Alexander Graham Bell has a variety of applications. It can be used to detect trace gases in gas mixtures at the parts-per-trillion level among other things. An optical beam chopped at an audio frequency goes through a gas cell. If it is absorbed, there's a pressure wave at the chopping frequency proportional to the absorption. If not, there isn't. Synchronous detection (e.g. lock in amplifiers) knock out any signal not at the chopping frequency. You can see even tiny signals when there is no background. Hearing aid microphones make excellent and inexpensive detectors so I think that the mics in modern phones would be comparable.
Contrast this with standard methods where one passes a light beam through a cell into a detector, looking for a small change in a large signal.
I don't understand how laser light, even from a small 5 mW laser pointer, can get the membrane to move. My first thought was that it wasn't moving and that the signal was from the photoelectric effect, but they have a section in the paper testing just that and it turns out it is the membrane moving. I doubt it's expanding from heat since it's such a low power and it reacts fast enough to simulate speech, but how does it work?
The thermal effect is one, where the membrane heats up and makes the surrounding air expand. The other is the light hitting photosensitive electronic components.
The heat has to go somewhere. The photoacoustic effect of the changing intensity of laser light causes the material of the case to generate sound which is coupled by direct mechanical contact to the mounted MEMs microphone.
That puzzles me, too. The thermal effect that drives a radiometer[1] should be too slow to transmit audio. Radiation pressure transfer from photons is very weak, but wouldn't have bandwidth problems. The frequency response of this effect does fall off with frequency, but not very fast; see fig. 6. in the paper.[2] There's still reasonable output at 3KHz.
Light pressure of sunlight is about 1mg/m^2. Or one nanogram per square millimeter, a reasonable size for a chip microphone. Will chip microphones respond to a nanogram of pressure?
This is amazing. I had no idea the "MEMS" microhones were light sensitive enough to exploit this. It's not the most far-fetched attack. If a well-funded government spy sees that there's an alexa in the room, why not tell it to make a phone call to a throw-away number and listen in?
Control over a voice assistant gives you a lot more than "just" a listening device in the room. It already has authorization for many actions inside the network. It's not that there are no other ways to achieve the same but this seems to be extremely cheap, easily exploitable, and with close to zero traceability.
> If a well-funded government spy sees that there's an alexa in the room, why not tell it to make a phone call to a throw-away number and listen in?
Because the more efficient way would be to just make use of the backdoors directly or force Amazon/Google to give you an eavesdropping API. All in the name of national security, of course.
> Light Commands is a vulnerability of MEMS microphones
I've always wondered how it's possible for smartphones to sound so good given that the microphone needs to fit in a few millimeters, how do they sound better than much bigger (cheap) microphones. Apparently the answer is MEMS microphones.
> Moreover, even if enabled, speaker recognition only verifies that the wake-up words (e.g., "Ok Google" or "Alexa") are said in the owner's voice, and not the rest of the command.
Holy shit, this might be even more astounding to me than the laser attack itself.
They explicitly mention that the effect seems unrelated to wavelength of incident light — would love to see a test with a UV or IR laser source to demonstrate that the primary downside of the attack is mitigatable (of course with additional expense and increased safety risk).
Based on this and the iPhone helium problem, it seems like the security and durability of MEMS devices needs to be examined carefully for their intended use case.
Contrast this with standard methods where one passes a light beam through a cell into a detector, looking for a small change in a large signal.
https://chem.libretexts.org/Bookshelves/Physical_and_Theoret...
Hats off to the Michigan team for this very clever (and unnerving) demonstration.