Great Circle Associates Firewalls
(June 1994)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: Cisco screening
From: MICHAEL NITTMANN <NITTMANN @ UWLAX . EDU>
Date: Thu, 2 Jun 94 15:17 CDT
To: firewalls @ greatcircle . com

To hide a network:

the point is not not to propagate routes, but to inhibit the 
router's public interface in responding to icmp echo requests. 

Maybe I misunderstood your term 'source filtering': I refer it to 
filtering packets that are sourced within a router. It is easier to 
apply such a filter on one place instead of having to repeat it on 
every interface that is there, and to maintain this when the 
interfaces are reconfigured in individual host.cfg files instead of 
one network.cfg file for all (sourced packets only if for all 
interfaces or all routers on the same interface, which is not really 
enforceable).
The facilitation is not to have to watch interface subcommands for 
common interdictions, but to put the filter 'behind' the router's 
packet source (from where stuff generated by the router itself is 
filtered globally for the whole box).

If you mean filtering of incoming packets instead of outgoing 
packets, there I agree with the Cisco opinion that it is not 
necessary. You can always write an equivalent filtering topology for 
only incoming filtering, or only outgoing filtering, giving the same 
results. Cisco's 'established' mechanism allows here for efficient 
ftp filtering. I am using it and it works: from outside you cannot 
ftp but from the inside there are no restrictions.

One thing I see in the messages is the problem of 'trust': my view 
is, that it is necessary to have a level of trust at least towards 
the people within the trusted domain. That trust is there anyways 
even without networks. There are many ways of transporting 
confidential information out of a company without using a traceable 
computer network. MOst companies do not employ sophisticated 
security towards the employees, so why then doing this on the 
network.
There is no defense against collaborating people that bring up a 
server with a certain socket and make thus available all data on 
their workstation, including anything that is mounted.
As is no defense against people that carry out drawings or documents 
'to work at home on it'. 
There are of course high security departments. But those have no 
direct network access anyways (that's where the employees' bags and 
pockets are searched before leaving the development/engineering 
area).

In my opinion, I would prefer to have someone use an IP network to 
compromise company confidentiality, since pretty much everything is 
traceable. This does not mean that every packet must be screened. 
Traffic statistics, suspicious traffic patterns, or all of a sudden 
lots of outgoing traffic from somewhere from where it is not useful, 
nor the habit, can trigger second level checks and, if proven 
necessary, tight supervision, up to tracing.

The difficulty of protecting agains consenting and cooperating 
employees in 'spy' cases should be obvious if you look at the best 
and secure facility you could think of: CIA.

Mike


Indexed By Date Previous: Re: Cisco software update?
From: "John P. Rouillard" <rouilj @ cs . umb . edu>
Next: Source routing with Ciscos: example to try out
From: MICHAEL NITTMANN <NITTMANN @ UWLAX . EDU>
Indexed By Thread Previous: Re: Cisco screening
From: Brent Chapman <brent @ mycroft . GreatCircle . COM>
Next: Re: Cisco screening
From: ted . doty @ nsco . network . com

Google
 
Search Internet Search www.greatcircle.com