Re: /dev/random is probably not
> - Linux? (I don't think so, If we have network and other I/O device
> such as keyboard, I thought that would be used, too.
> but I want confirmation from people in the know.)
It's been a while since I looked at the /dev/random design on Linux (probably
the early 2.4 days), however one thing that was quite clear was that they did
not use any network I/O as entropy sources because an attacker, particularly
one that already had control of other machines on the same LAN segment, could
have a high degree of control over that source. I would be most interested if
that has changed since the last time I looked at it.
> OpenBSD, FreeBSD, NetBSD and the like, and of course
Checking the /dev/random manpage on Darwin, it indicates that entropy is
input from the system "Security Server", which uses "kernel jitter".
Unfortunately, a quick check did not reveal exactly what the source of this
kernel jitter is. Never-the-less, the manpage does indicate that this
/dev/random design is from FreeBSD and likely shared with other BSDs.
> Windows family OSs.
All I can observe here is that F-secure SSH still (at least the most recent
version I've used) collects its own entropy when running on Win2K, which
indicates to me that either they want to operate the same on all Windows
versions (as memory serves, Win95/98 does not have a RNG), or that Win2k does
not have a suitable RNG.
> One of these days, on desktop PCs,
> we could add the reading of diode used for measuring
> CPU temperature to the mix of
> entropy source. (Of course, we need a good source of
> `entropy' to begin with, and adding another source such
> as diode is a good thing IMHO.)
> And maybe the fan rotation/speed, too. I found that
> they change constantly on my PC!
You would only want to use one or the other, since the fan rotation is a
function of the CPU temperature measurement -- if you used both you would
essentially be entering the same measurement into the RNG twice, which isn't
very random.
> Some of these CPU-bound devices may have
> implications when we have a dual core CPU.
> Reading of such device by one thread may be
> highly predictable by another thread running on the
> CPU chip.
Indeed -- certainly the recent advisory regarding information leakage through
the cache between threads on multi-core CPUs (CVN: CAN-2005-0109) indicates
that we're starting to find problems of this nature already.
Cheers,
Terry
#include <stddisclaim.h>