Re: Randomness Attacks Against PHP Applications

[Raphael Geissert <geissert () debian org>]

On Monday 17 September 2012 10:36:46 Josh Bressers wrote:
> > On Wed, Aug 22, 2012 at 02:31:07PM +0400, Solar Designer wrote:
> > Maybe these PoCs will help convince someone.

Just a note regarding the sessionid case: IIRC since 5.4 
session.entropy_length is set to, erm, 32 (bytes.)
Basically it appends N bytes from /dev/urandom to the other input for the 
digest and then it is computed.
(why 32 bytes, and why still use md5 by default, well...)

> I'm skeptical they will. I've been doing a lot of work for the past year
> on various proactive security efforts. I keep coming back to two basic
> things.
[...]
> Has anyone tried to talk to them about this further to see if
> the issue is they don't understand, or are they being stubborn?

I think the main problem is education. For instance, there is no word about 
mt_rand not being suitable for criptographic pourposes (much less what that 
means.)
Sure, searching for "crypt" in the page shows a few comments saying that it 
isn't suitable, but:
a) there are far more "encryption functions", "random password generators", 
and similar stuff in the comments than those that do mention its weaknesses. 
b) the official documentation itself doesn't say a word. It should say it loud 
and clear.

Comments should also be moderated. Many examples available as comments in 
the documentation are incorrect.

Now, pointing it out is easy, but somebody has to actually do the work. 
*That* is another issue.

Cheers,
-- 
Raphael Geissert - Debian Developer
www.debian.org - get.debian.net