[Cryptography] Any Electrical Engineers here who know about noise filtering?

[Henry Baker]

At 02:34 PM 1/2/2016, Natanael wrote:
>For the latter (which I guess you really mean), the pad provably brings up the noise to a mathematically identical level for all bits, there is no signal leak.  *It isn't addictive noise* like with regular signals such as electromagnetism or audio.  You can't subtract anything based on any filter, because everything is indistinguishable from random.  The whole point is that OTP eliminates correlations.

Ok, let me take a whack at a proof.

The noise has uniform spectral density, so the best matched filter would be a pure flat bandpass filter -- i.e., the identity filter.

But the Fourier transform of a flat filter (in our circular convolution world) is a single impulse of unknown phase.

Convolving that impulse simply samples/reconstructs the original signal.

So the best-matched filter improves the signal not a whit (!?!).