It seems you don't know the first thing about the topic you are talking about. Side-channel attacks are very real. I'm working in a lab which focuses precisely on cryptosystems' implementation security, and I for instance saw someone extract the secret key of an AES running on a last gen smartphone (with multiple cores etc. as you describe).
The attack was using an antenna (which listened to electromagnetic emanations rather than sound, but that's a detail, it is also possible to do that with power consumption, timing…) which was simply put near the screen of the phone. That was not supposed to be very impressive, it was during an exhibition just as a fun demo to attract people to the stand which did it.
And moreover, about what you say in your criticism, cryptographic computations often take place on embedded systems like a credit card where you have less noise and slower computations, which makes it even easier to perform this kind of attacks.
We are talking about acoustics, sub 100khz signals. They are not talking about catching emanations from sound codecs or key strokes they are talking about CPU instructions for crying out loud. There is certain loss of information when we are talking about catching x86 instructions happening in 3Ghz. Paper does not really explain this.
Comments
It seems you don't know the first thing about the topic you are talking about. Side-channel attacks are very real. I'm working in a lab which focuses precisely on cryptosystems' implementation security, and I for instance saw someone extract the secret key of an AES running on a last gen smartphone (with multiple cores etc. as you describe). The attack was using an antenna (which listened to electromagnetic emanations rather than sound, but that's a detail, it is also possible to do that with power consumption, timing…) which was simply put near the screen of the phone. That was not supposed to be very impressive, it was during an exhibition just as a fun demo to attract people to the stand which did it.
And moreover, about what you say in your criticism, cryptographic computations often take place on embedded systems like a credit card where you have less noise and slower computations, which makes it even easier to perform this kind of attacks.
We are talking about acoustics, sub 100khz signals. They are not talking about catching emanations from sound codecs or key strokes they are talking about CPU instructions for crying out loud. There is certain loss of information when we are talking about catching x86 instructions happening in 3Ghz. Paper does not really explain this.