mailing list archives
Re: CVE for OpenBSD random() bug?
From: Kurt Seifried <kseifried () redhat com>
Date: Fri, 23 Mar 2012 12:55:10 -0600
On 03/22/2012 07:24 AM, Todd C. Miller wrote:
It would seem this fits into the "weaker then advertised" class of
security problem. Thoughts/comments (anyone strongly against this)?
Since random(3) is not a cryptographically secure random function
I'm not sure that is makes sense to assign a CVE.
I suppose it really depends on the likelihood of someone calling
srandom(0); I don't know why anyone would do that on purpose. If
you must use random(3) instead of something stronger like arc4random(3),
it is possible to seed the PRNG via /dev/arandom using srandomdev(3)
or set the seed state manually via initstate(3), both of which
provide more than just 32 bits of seed data.
I guess if no-one thought it would be used insecurely it wouldn't need
to be fixed ;). Also not all security uses of randomness are strictly
crypto related (e.g. array seeds to defeat HashDoS, etc.).
Please use CVE-2012-1577 for this issue.
Kurt Seifried Red Hat Security Response Team (SRT)