Video retraces as a source of entropy...

["Deranged Mutant" <WlkngOwl@UNiX.asb.com>]

Quickie blurb on using video card retraces as a source of entropy...

I've done some brief testing in the last couple of days on using the 
timing drift between video retrace events as a source of randomness.
It seems comparable to truerand() spinners that check the system's 
timer ticks [which makes me leary of relying on it since it
has similar strange attractors when plotted in a noise sphere, but
that's another post...].

Assuming one trusts truerand spinners, this method could have some 
advantages over a 'pure software' method, since the video controller
(and other hardware controllers which could be adapted to this) runs 
in 'parallel' [although it's liable to the samefluctuations in current or 
memory-access and other interfaces with the main system, or possibly 
tempest attacks for the paranoid...].

In pseudo-C:

int retrace(void) { // test for video retrace
#ifdef __MSDOS__
  return (port[0x3da] & 8); // Some VGA, maybe EGA cards
#else
  // your OS here
#endif
}

  [..]
   x = 0;
  while (!retrace()) x++;

I've tested it as  standalone routines (in Pascal and assembler)  as well as
a hook to the DOS idle in the background [See note about strange
atteactors above].  It also seems to work while in Windows (but not
OS/2?!), which is an advantage over using the system's microsecond
timer alarm (which Win3 takes over).

Comments? (other than "truerand is an oxymoron"....)

Rob. 

---
Send a blank message with the subject "send pgp-key"
to <WlkngOwl@unix.asb.com> for a copy of my PGP key.