Using /dev/random for PGP key generation? Be Wary

[Bill Frantz]

At 10:56 AM 2/12/96 -0500, rngaugp at alpha.c2.org wrote:
>...
>I am unsure about using my modification, together with these drivers
>that are not connected to a real hardware RNG. In what way would the use
>of these drivers' methods of gathering entropy be superior to PGP's
>method of getting entropy from keyboard timing? If you choose to do
>something like this, you should think carefully and make a careful study
>of the code.

If there are no common-mode sources, xoring two streams will not reduce the
entropy.  If you use PGP's keyboard timings for one stream, and (e.g.) disk
drive randomness for the other, the output of the xor of the two streams
should have at least as much entropy as the best of the two.

However, I would be worried if /dev/random and PGP were both using keyboard
timings to generate entropy.

Bill