[7264] in cryptography@c2.net mail archive
Re: random seed generation without user interaction?
daemon@ATHENA.MIT.EDU (Don Davis)
Tue Jun 6 11:28:26 2000
Date: Tue, 6 Jun 2000 08:59:37 -0400 (EDT)
Message-Id: <l031107000000863c4d36@[208.192.101.169]>
In-Reply-To: <200006060208.TAA29744@breakaway.Stanford.EDU>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: cryptography@c2.net
From: Don Davis <dtd@world.std.com>
> what methods do folks currently use (on NT and unix)
> to generate a random seed ...
mr. hodges,
solaris has a good trng product called
cryptorand. i've reviewed its internals
closely. cryptorand works by hashing
kernel memory. the pointers in kernel
memory get shuffled constantly by external
I/O, so /dev/kmem does contain good
randomness. cryptorand is very careful
and conservative about not delivering
too many bits.
linux' /dev/random hashes various load
averages with various sources of i/o
timing. i believe it is sound.
nt uses capi's generate_random_seed() call,
which mostly relies on static configuration
data, like the host's name & ip address.
if i remember correctly, it may hash the
load average in, too.
- don davis, boston
-
