Re: entropy masking (was Re: Multiple spinners as sources of entropy?)

In a message dated 96-03-13 18:35:51 EST, Matt Blaze writes:

>I would go even further than this.  I wouldn't trust ANY
>environmentally-based random source (cycle spinner, keyboard
>timer, disk noise, whatever) against adversaries on the
>same system.

In DOS, the keyboard, mouse, and disk drives run on interrupts, not timers.
 If you use a timer as a spinner, such as the Windows GetCurrentTime()
function, (I MSec. resolution) and check its value each time a key is pressed
and released, It would appear that the results should be quite random,
because the keystroke is not processed in conjunction with the timer, but
rather whenever the interrupt occurs.  My tests in this area indicate that
any 8 bit value can be achieved with this method, with a fairly uniform
distribution.  My tests are not thorough (I haven't sat down and typed for 2
hours to test the distribution of the output), but results look reasonably
good so far.

Jonathan Wienke

On Wed, 13 Mar 1996 23:47:58 -0500, you wrote:

>In DOS, the keyboard, mouse, and disk drives run on interrupts, not timers.
> If you use a timer as a spinner, such as the Windows GetCurrentTime()
>function, (I MSec. resolution) and check its value each time a key is pressed

[...]

  I'd also recommend using mouse clicks. I think that most windows users
(going from personal experience) spend much more time clicking through
messages than typingon the keyboard. A combination of several events should
provide 'even better' randomness.

   Brian


------- <blane@aa.net> -------------------- <http://www.aa.net/~blane> -------
  Embedded Systems Programmer, EET Student, Interactive Fiction author (RSN!)
==============  11 99 3D DB 63 4D 0B 22  15 DC 5A 12 71 DE EE 36  ============