On 5 Apr 2002, at 19:49, Bill Royds wrote:

> That is a reasonable  fit to the term Paul and I were using, if we
> consider a screening router a type of firewall, and the one used
> in Chapman and Zwicky. Unfortunately (and the reason I asked for
> clarification in the original post) others have used the term for
> a segment connected separately to the firewall as in 
> 
> 
> Internet -----[firewall]-------internal
>                   |
>                DMZ?  
> 
> This DMZ? has different security characteristics from the others.
> It routes differently and the use of the term DMZ for it confuses
> matters rather than clarifies them. 

  Barring catastrophic defects in the firewall code itself, the 
primary difference between this and

Internet --[firewall1]--DMZ?--[firewall2]--internal

is that in the separately-connected segment case, a compromised 
server in the DMZ? network has *no* possibility of sniffing traffic 
between the internal net and the outside.  (In the two-firewall 
scenario, the opportunity for this can be reduced by using a switch.)
  The key point is that the DMZ? segment and the internal segment are 
separated from each other by a security-policy-enforcement boundary.  
Whether this boundary, and the boundary between the DMZ? and the 
Internet, are implemented by the same device is topologically 
irrelevant.

  What's NOT irrelevant is whether the security policy boundary 
between the internal network and the internet is explicitly 
specified, or arises out of composition of the internal/DMZ? and 
DMZ?/Internet boundaries.  A debatable question is whether the 
composition case is really "more secure", or merely "harder to manage 
and, as a consequence, harder to accidentally leave hanging open".

  (It has been suggested that there is some benefit to compositing 
two different firewall technologies, but the usually recommendation 
in that case is to use a packet filter for [firewall1] and an 
application proxy for [firewall2] FOR PERFORMANCE REASONS.  It seems 
to me that if our primary concern is defensive security (as opposed 
to content monitoring...), then [firewall1] is where the application 
proxy is needed, and that a mix of caching and redundancy is the 
proper way to address the performance consequences.)

  [Neither of these scenarios is a very good parallel to the original 
use of the term to refer to real estate around the Korean cease-fire 
line.  If it weren't so late in the day, I'd favor a push to expunge 
the term "DMZ" entirely from the Network Security vocabulary.
  I also really dislike router appliance vendors who've begun using 
"DMZ" to mean "static NAT to a server on the internal network".  If 
you don't have a policy boundary between it and the internal network, 
it's not a "DMZ" by *any* sane definition.]

Dave Gillett




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

Reply via email to