On Mon, Apr 15, 2002 at 08:06:39AM -0500, Noonan, Wesley wrote:
> 
> position that seems to be so prevalent on this list. I think it is bad
> practice to make such blanket generalizations.
>

OK... lessee what happens if we take another approach to this whilst
holding this thought in your head: "trusting that the layer above
below you will do the correct thing with invalid data leads to
security breaches", consider the following:

a) VLANs are implemented by putting data into an otherwise unused part
   of the IP header.  Anyone can easily craft a packet with the right
   data in the right place given the right tools and access.

b) A switch which receives a packet with VLAN information on it on a
   non-trunk port should treat the packet as bad (well IMHO) and
   either drop the packet or scrub the invalid information.  The
   reality is that the behaviour will be "implementation defined"

Given these two points, can it be considered wise to trust the
implementation of VLANs on a particular switch?  Well, perhaps, if you
can verify the switch behaviour by extensive testing but what if you
miss something (this sounds like FUD, I know) - just take the example
of overlapping packet fragments being used to bypass firewall checks
not that long ago, or what if the tested switch fails and you get a
new one - do you repeat your tests because the ASIC's in the switch
may have changed?  Given the cost vs risk it is simpler and more cost
effective just to get two separate switches.

> 
> Bingo!!! This is what I have been trying to get across. Everything has it's
> place, and in many cases VLANs have their place as an aspect of a perimeter
> design. Is it the "best" design? Maybe, maybe not. Can it be the best design
> for the circumstances and requirements? Absolutely.
>

ok - so, you are saying that multple switches does not scale well but
  look at what you are suggesting, something that you are saying is
  pretty much unacceptable in a single host situation suddenly becomes
  ok in a large scale deployment?  Where you have much more complexity
  and exposure?  I am not so certain that this is a good thing - I am
  not suggesting a "damn the torpedoes" solution of a small switch for
  every connection but one must be wary of trying to make a lame duck
  fly simply by expanding the size of the duck.
  
> 
> Certainly not. However, chasing boogeymen isn't a good practice
>  either.
>

Hmmmm there have been times when today's boogeyman is tomorrow's
nightmare.  One must not be complacent.

> As
> security people, and in opinions expressed on this list many times, people
> seem to forget that business still has to get done. Security is, and should
> be, secondary to making money. This is, after all, business and capitalism.
> Security policies and practices that prevent business are BAD.
>

Sorry, no.  A security policy is there to help you STAY in business.
A BAD security policy stops you doing what you need to do.  Security
is not convenient.  It is not fun.  I would much prefer not having to
manage a firewall, update virus patterns, lock things down but because
of a whole load of twerps we are forced to do this or suffer having
the resources we manage soaked up by said twerps serving up porn or
warez to their community - there is no money in doing that so we do
our best to prevent it happening.  Security is a fine balance between
getting what you need done without exposing yourself to being used for
some unprofitable purpose.
  
> 
> Agreed. Again, this is the point that I was trying to make. It's not cookie
> cutter. There are a LOT of variables to weigh, and I just think it is bad
> practice to make statements like I have seen from others on this list. 
> 

It is not a cookie cutter but proposing to use VLAN's to enforce a
security perimeter seems to place a large amount of trust in the fact
that the switch will deal with "erroneous" packets correctly[1] AND that
the switch has been configured correctly[2] AND that any replacement
switch will be similarly configured.

[1] Actually, there were some reports on bugtraq that people had
managed to get packets to jump vlans on 3com and cisco switches.

[2] Some 3com enterprise switches default (well, used to at least) to
"open mode" which means that they do not constrain the VLAN traffic to
ports that are members of that VLAN, you needed to put the switch into
"closed mode" to fix this.

-- 
Brett Lymn
_______________________________________________
Firewalls mailing list
[EMAIL PROTECTED]
http://lists.gnac.net/mailman/listinfo/firewalls

Reply via email to