You are correct, it cannot. I've spent a good portion of my career working on networks which are CLASSIFIED or higher. There are very strong restrictions on simultaneous interconnection between systems of varying classifications. In most cases, complete -physical- isolation is required, although waivers may be obtained on a case-by-case basis. The enforcement of these policies of course varies between agencies, bureaus, and departments. I remember just a year-and-a-half ago installing fibre switchboxes on some USAF nodes which at various times needed to have connectivity to networks of varying security levels. This is slowly changing at the lower levels due to the advances in proxying technology, etc., but when you reach TS/SCI or SAP/SAR it's a different story, and understandably so. As a side note, the reason the first version of Microsoft Exchange, 4.0, was so late was due in large part to the fact that Microsoft wanted to bid on the Defense Messaging System, or DMS contract. DMS supports all levels of classification and types from UNCLASSIFIED all the way through FLASH, CRITIC, etc. The powers that be understand that an application-level service like messaging has to be able to cross such boundaries, but only with strong rules which prevent the dissemination of information of higher sensitivity to those who aren't cleared for same. To the best of my knowledge, there are no networks in DoD, the armed forces, or various other arms of the U.S. government which are ranked as SECRET or above which have any form of Internet connectivity at anything less than two removes. One of the reasons that it's so hard for them to come up with a coherent and workable overall information security policy is because of the (very necessary, more so than the typical slashdotter tends to believe) compartmentalization of information - and hence, IT resources - within various organizations which deal with sensitive information and issues. "Greg A. Woods" wrote:

[ On Sunday, July 9, 2000 at 08:22:46 (-0700), Roeland M.J. Meyer wrote: ]

...

Subject: RE: RBL-type BGP service for known rogue networks?

In many organizations, a system isn't considered secure unless port 22 is blocked, at the firewall. It is, after all, the secure port, that must mean that you have to block it to be secure, right?

Yes, that's exactly right, but not for the reasons you imply.

If the primary concern of a security policy is that covert channels must be prevented then it is absolutely mandatory that port-22 be blocked since it is by definition a covert channel.

However having any kind of Internet connection, proxied or not, into a site where sensitive information must not be allowed to be leaked is in effect a violation of the policy.

Unfortunately we're rapidly approaching (if we're not already there) a state of affairs where it is impossible to technically prevent inbound and outbound covert channels wherever people are required to interact in a privileged way with security sensitive systems. A paper given at last year's ACM New Security Paradigms Workshop by Dean Povey ("Optomistic Security: A New Access Control Paradigm") suggests that it might be better to adopt the view that security officers should "Make the users ask forgivness not permission." Whether this paradigm can successfully be delployed in top secret (or higher) environments or not is yet to be discussed. I suspect it can but then I'm not an expert in traditional forms of high security.

-- Greg A. Woods

+1 416 218-0098 VE3TCP <gwoods@acm.org> <robohack!woods> Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>

-- ----------------------------------------------------------- Roland Dobbins <mordant@gothik.org> // 818.535.5024 voice One of the surest signs of the philistine is his reverence for the superior tastes of those who put him down. -- Pauline Kael