[Cryptography] Fwd: [IP] 'We cannot trust' Intel and Via's chip-based crypto, FreeBSD developers say
Bill Cox
waywardgeek at gmail.com
Tue Dec 10 18:57:31 EST 2013
I have to take back my criticism of Intel's RNG. I got my sims working
for a version of their architecture in .35u CMOS, and it's simply better
than my "Infinite Noise Multiplier". It's probably the best true random
noise generator ever. I still don't like how their schematic is seems
highly sensitive to supply noise, but we don't know what the actual
circuit looks like. Intel hasn't told us.
So, I'm going to modify it a bit to use the resistors available on my
chip and reduce the caps, fix the supply sensitivity, and I think I can
run 16 of these things in parallel at 100-200MHz on the tiny .35u CMOS
chip I'm designing. I'll spit out the raw waveforms from the inverters,
buffered once, through 16 "analog" pins, so there wont be any fear
(hopefully) that I'm cooking the data on-chip, before you can see it,
and I'll open-source the schematics. If there's a circuit that can
consume all 1.6Gbit/sec of this raw data, have fun with it! On the
digital side, I'll XOR bits together to get the bandwidth down to
something reasonable, which I can send over USB, and provide a simple
Linux driver.
This thing will definitely put out RF, but since I'm making the raw data
available at the pins, should I care? By the way, this is just a
for-fun project at work. I get to do a free chip design :-)
More information about the cryptography
mailing list