Re: Subtract-and-carry PRNGs


Augustus SFX van Dusen wrote:
Subtract-and-carry PRNGs are attractive because of their simplicity, and
because of the fact that they tend to have extremely long periods.
However, I understand that at least some implementations show poor
statistical randomness properties. My question is, Are subtract-and-carry
PRNGs in general prone to have (relatively) poor statistical randomness
properties? If not, are they cryptographically safe?

If they're anything like a lagged fibonacci generator they're totally
linear. So no, you can't use them directly for cryptographic
purposes.

If you want a secure PRNG you should look into something along the
lines of Yarrow or Fortuna.

Tom

.