Re: Collecting entropy in headless systems
Ralf-Philipp Weinmann <ralf@fimaluka.org> writes:
Have a look at Havege [1,2]:
[1] HAVEGE [ HArdware Volatile Entropy Gathering and Expansion http://www.irisa.fr/caps/projects/hipsor/index.php
I've had a quick look, it just reads something like the Pentium TSC in a loop and, if the delta is more than a certain value, decides that an interrupt has occurred and takes the result as a new entropy value. There's a lot of handwaving, and some more stuff involving (hopefully) the nondeterminism resulting from a CPU cache-thrashing algorithm. It's just another in a long series of cool-but-unverifiable ideas for gathering info from a CPU, the first being the clock-skew mechanism from the early 1990s (time a fast clock using a slow clock, this predates the Pentium so the original version used a fast loop sampled from the 18.2 Hz PC clock). If you really want do go down this path, use Matt Blaze's TrueRand, circa 1995 (and unpatented). Peter.
participants (1)
-
pgut001@cs.auckland.ac.nz