http://tools.ietf.org/html/draft-ietf-opsawg-firewalls-01#section-3

I'm copying the list as much as anything to generate discussion and make sure 
we get the right stuff into the draft.

Thinking about the section on policies, it seems a little disjoint. I find the 
use of the words "newer" and "modern" a little dismaying; to begin with, I'm 
not sure that they are newer or even modern; anomaly-based and signature-based 
intrusion detection has been around for rather a while. The terminology also in 
effect deprecates the zone-based policy that is in fact pretty common, and is 
specified in RFC 6092. The notion that firewalls significantly change with the 
deployment of IPv6 is also, while cherished in some quarters, IMHO naive; if 
one's corporate policy is that firewalls should implement a certain information 
security policy with IPv4, I almost guarantee that they will do the same for 
IPv6 or any other communication service. I think I would prefer to have an 
actual enumeration and discussion of policies, and leave opinions on them to 
the "recommendation" sections.

I suspect that there are four broad classes of policy in the field. Happy to 
hear that there are others. I'm going to state these in Cisco language because 
I am comfortable with it; if someone else likes other language, I don't have a 
problem with that - send text.

One is what we call a context-based or zone-based defense. My enterprise 
network includes some set of systems and connectivity, and I draw a physical or 
topological perimeter around it. The policy may permit peers outside the domain 
to access a defined set of systems for a defined set of services - web accesses 
to the web servers, mail to the mail servers, and so on. Apart from that which 
is explicitly permitted to be originated from outside the domain, the only 
sessions permitted to cross the firewall are those that originate from inside 
the perimeter. To be honest, I think of this perimeter as the CIO peeing on the 
trees around his territory - he understands that this is prophylactic, and that 
it provides limited defenses, and uses it as much as anything to take control 
of his perimeter.

Another is perhaps best described as an open defense. Ingress filtering 
(filtering of messages based on routing information) is a classic example of 
this. Even if I have no policy regarding applications crossing my perimeter, I 
very likely do not permit traffic from outside of my network to provoke attacks 
within my network. An example of such an attack is someone ending pings to 
various systems in my network with a source address within, resulting in a ddos 
on a particular address within my network.

A third, which Warren Kumari commented on, is a policy to interrupt observed 
attacks. One "observes" an attack using an anomaly-based or signature-based 
algorithm, perhaps both, and in the event that such an attack is identified, 
one asks someone upstream to block an identified set of traffic.

A fourth is what we call a role-based defense. In a data center, for example, I 
may only permit stated sets of tenants to talk with each other. A single 
administration within the data center may have multiple tenants - storage 
systems in one, computational resources in another, and so on. Within a 
corporate network, one frequently finds what look like perimeters around 
departments like HR or Finance, with anyone in the department being able to 
access a wide set of resources, but folks outside the department only being 
able to access a limited set of resources. The screwy thing is that "folks in 
the department" are one person or one office in each of a number of locations. 
So the "perimeter" is a set of persons or their computational proxies rather 
than a physical or topological domain.
_______________________________________________
OPSAWG mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/opsawg

Reply via email to