These days, I'm applying lots of DSP thinking to other kinds of signals to
understand privacy leaks and ways around them. (It turns out that DSP
thinking is quite useful when it comes to thinking about what metadata is
being leaked and how to reduce that leakage).
True random is a common need for cryptography, so I've been thinking about how
best to feed sensor/mic noise to the kernel's random source. One idea is to
only use the ADC/accelerometer data when the RMS amp is low, so the signal
much more likely to be all noise.
Someone else proposed this idea, and I was wondering whether it makes sense:
> How about just keep the least significant bit from each sensor source, and skip any thresholding
> based on on amplitude or signal energy? Then for multiple sources, perhaps
just XOR the bit
> streams together.
That reminds me, the simple Pd question I have is: anyone know of an XOR tilde
object?
.hc