[Octave-bug-tracker] [bug #58800] BIST for rng sometimes fails

[Michael Leitner]

Follow-up Comment #18, bug #58800 (project octave):

The Mersenne Twister has a practically infinite period (in the context of
computing power available today or in the foreseeable future) of 2^19937−1,
and it of course has a corresponding number of states. That means, if the
seeds are different, the generated pseudo-random number streams will have no
detectable relation. However, if we would only have 2^16 different possible
seeds, we would have only 2^16 different streams. That's too low, you can
easily enumerate that. Even 2^32 is too small. I think that we should have at
least 2^64 different seeds. 

And further, if it was only for linux with /dev/urandom, I see nothing that
would speak against directly using that: when I request 4*625 bytes from
/dev/urandom, /proc/sys/kernel/random/entropy_avail shows increasing rather
than decreasing content of true entropy (which, if it fell to zero, would make
/dev/random block). Of course, I do that from the command line, and the exact
time when I do the call constitutes a source of entropy, while it could be
that calling that in a program is considered as deterministic and does not
increase entropy. But in any case, I would say /dev/urandom drains the
available entropy only very slowly. However, I do not know about other
systems.

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?58800>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/