← Back to context

Comment by londons_explore

17 hours ago

> voice communication that only your device can hear.

This is fairly straightforward - you have the device spew out noise with similar characteristics to human speech (ie. random overlapping syllables in the speaker's voice). Take a recording then subtract the random syllables.

Only your device can do the subtraction, because only your device knows the waveform it transmitted.

Obviously in a room with lots of reverb this will be a bit harder, since you will also need to subtract the reflection of what was transmitted with a room profile and deal with the phone moving in the room, but it sounds far from impossible.

Countermeasure: set up four microphones some distance apart, use autocorrelation to pinpoint the sound sources, and then isolate them, recovering the "masked" speech. The countercountermeasure would be to fully surround your mouth and vocal tract with an active noise cancelling system and then produce noise (to push whatever little sound gets through far below the noise floor: the signal is unpredictable enough that you can't use averaging techniques to recover it). The countercountercountermeasure would be to use a camera in the radio band to look at the vocal tract directly, using the phone as a light source, and recover the phonemes that way. The countercountercountercountermeasure would be to construct an isolated box… at which point you're no longer having a voice call in public: you have a portable privacy booth.

> you have the device spew out noise with similar characteristics to human speech

Surely this only works if you're using the phone as a speakerphone (and are therefore almost certainly being an arsehole in public[0])?

[0] Because if it was an actual speakerphone situation, hiding your voice would be stupid.