|Main Archive Page > Month Archives > full-disclosure-uk archives|
-----BEGIN PGP SIGNED MESSAGE-----
On Mon, 19 May 2008 13:51:29 -0400, Garrett M. Groff wrote:
> Generating pseudo-random numbers isn't hard given a good API, but
> writing that API is non-trivial (assuming you want high entropy/low
> predictability). And, apparently, screwing up that API is also very
Generating real pseudo-random streams is a hard problem which is way more than what people can handle. Usually, PRNGs are composed of various periodic elements which, in the end, all combined produce a repeating stream of pseudo-random numbers. OpenSSL uses a modified MAC for this as a state machine and extracts some state bits as random stream on every access.
We're not debating the PRNG itself here but the _seed_. OpenSSL supports various ways to influence the state of the PRNG at various stages by XORing in new material; however, Debian chose to only support pre-seeding with uninitialized memory and the Process ID. With the elimination of the uninitialized memory as seed, the seed for the MAC was entirely comprised by the PID.
So we're not debating a weakness in the PRNG here at all, which is a _very_ delicate subject.
Solutions Systematiques Tel:+41 61 333 80 33 GŁterstrasse 86 Fax:+41 61 383 14 67 4053 Basel Web:www.sygroup.ch firstname.lastname@example.org
-----BEGIN PGP SIGNATURE-----
Version: PGP Desktop 9.6.2 (Build 2014) - not licensed for commercial use: www.pgp.com
-----END PGP SIGNATURE-----