Re: Patching

[Ansgar -59cobalt- Wiechers <bugtraq () planetcobalt net>]

On 2003-10-21 Alessandro Bottonelli wrote:
> On Tuesday 21 October 2003 10:33, Ansgar -59cobalt- Wiechers wrote:
> > On 2003-10-20 Alessandro Bottonelli wrote:
> > > Hmmmm. I am not convinced yet that all this makes sense from a
> > > "wider" security perspective. Must a vulnerability / hole be known
> > > to be a risk?
> >
> > Yes.
>
> The more I think about it, the more I do not agree. Security is
> availability, confidentiality and integrity, isn't it? An unknown hole
> / vulnerability can still hit you hard (data loss, data integrity,
> system availability to name a few instances). Humans may not know
> about such vulnerability but systems run that code, and if the code is
> flawed, systems do not need humans to fail or to behave incorrectly
> from a security perspective. 

Availability, confidentiality and integrity are separate issues that
have to be addressed in different ways. When talking about security (at
least on this list) the majority is referring to confidentiality and
partly integrity (to the extent that data isn't manipulated by
unauthorized persons). At least that's my perception. Please correct me
if I'm wrong.

You are right that might affect availability and/or integrity and that
those issues need to be taken into consideration, but AFAIK they are
usually not considered _security_ risks.

[...]
> Was the price of closing a known hole that maybe someone one day might
> have exploited (and maybe I might have had another option for
> proctecting my systems) worth a failed Disaster Recovery?

Deal with that problem when you run into it. That's what you test
patches for. There is no point in avoiding a patch just because it
*might* break something.

If it breaks something you will have to make a decision whether to apply
it (and forfeit on some functionality) or not (and face the risk of
getting 0wnzed). That decision can only be made for each individual
incident.

If anyone happens to have a golden rule here, I'd like to know too ;)

> I am not saying patching is evil, but is dawning on me the idea that
> is not "necessarily" good, or in other words its worthness is not
> axiomatic.

It may rise other problems, but it gets you rid of a security breach
that is known to the world. That's the only point I was referring to.

> The list suggested a testbed system should be used for testing patches
> before going onto production systems. This would be a good step
> forward in making patches less dangerous, yet many organizations (or
> at least most of those I deal with) cannot (or do not want to) afford
> such luxury which requires a duplicate system, time and human
> resources (and even then I wonder how thorough and reliable a test
> would be on a non-production system, probably not fully interconnected
> with the whole infrastructure).

There has been a discussion on this a while ago, and there were some
valuable suggestions (e.g. using VMware). You might want to take a look
into the list's archive.

Regards
Ansgar Wiechers

---------------------------------------------------------------------------
Visual & Easy-to-use are not words that you think of when talking about 
network analyzers. Are you sick of the three window text decodes? Download ClearSight Network's Analyzer and see a new 
network analysis tool that 
makes the complex - easy
http://www.securityfocus.com/sponsor/ClearSightNetworks_security-basics_031021
----------------------------------------------------------------------------