CTO Chronicles

NAC now? NAC later? How about both?

Since it's sum-sum-summertime, I'm inevitably reminded of afternoons spent at the swimming pool, which in turn (of course) reminds me of candy.  Now, everyone had their own favorite candy back then.  Some would continue to insist on candy that, in the space of about 3 seconds, was going to melt in the heat that is a trademark of Texas summers.  Others would go for a more heat-resistant candy that, while far less delicious, at least wouldn't melt away like a wet wicked witch.  My own favorite was Now and Later.  Marketing delayed gratification to kids is always a tricky business; however, these candies were a stroke of genius.  Eat them now.  Save them for later.  They're a tasty treat either way.

I was reminded of all of this last week, reading Tim Greene's article on the Gartner Marketscope report.  Greene summarizes Gartner's viewpoint that even initial NAC deployments must take into account the organization's larger strategic vision of network security, endpoint compliance, etc.  Certainly, we agree with that.  However, the article also seems to pose a choice to organizations between an "overlay" solution that may have a short shelf life (say, a Snickers bar), and a more broadly integrated infrastructure-based solution that's not quite there yet (those fun dip things that had the inedible edible stick).

Any pure-play (overlay, whatever you want to call it) NAC vendor that cannot articulate a vision for how their solution fits into a larger overall framework is going to melt away very quickly.  And even before it melts away altogether, organizations are likely to get disgusted with what it's turning into and simply toss it.  By contrast, current pure-play vendors that do have that vision are like the Now and Later candy.  They provide a good firm texture now, but are pliable enough later to mold into your general network security strategy.

So, I say the choice is not as stark as portrayed in Tim's article, either for customers or vendors.  You can get your NAC snack now, knowing that even after some cannonballs and back flips, you'll still have something that is not just relevant, but delicious.

More Musings on MS and Managed

Interop remained all atwitter about MS NAP, and the press articles continue.  Joel Snyder had a webcast on Network World, where MS NAP was the main topic.  We've got a whitepaper up on the same topic, and Alan Shimmel continues to post as well.  There are nits that I could pick with just about everybody (e.g., most recent Cisco gear is perfectly capable of running per port ACL's), but I want to try to stay as high level as possible.  Here are the big-picture things I still see missing from the discussion:

Device Discovery

Whatever governance coverage NAP brings (Greenstein's comments are well taken here), it brings that governance for Vista and XP 3 machines owned and managed by the organization.  That leaves an extremely broad swath of devices, both specialized and general computing, outside the NAP umbrella.  Snyder mentions using MAC based authentication and external device profilers to help; but for many organizations with highly dynamic networking environments, just getting the list of printer MAC addresses (much less badge readers, PDA's, security cameras, the list goes on) is a daunting task.  And having an(other) external tool map and profile devices strikes me as a science experiment run amok.  It seems to me that it is critical to have a built-in device discovery and profiling mechanism.  It goes without saying that the discovery mechanism must also be impervious to host based firewalls, transparent to the physical connection, and indifferent to address assignment.

Post Admission

Just when I thought post admission security was in, they pulled it back out.  What's up with this seriously?  Gartner, Forrester, Frost & Sullivan, Tim Greene over at NWW all seem to agree that NAC encompasses (or at least should encompass) policy enforcement throughout the endpoint's network lifecycle.  The TCG announced a new specification last week aimed specifically at post-admission security.  Yet the bulk of Snyder's talk completely ignores post-admission monitoring and enforcement.  Even the perennial and predictable "blah-blah is dead" post from Stiennon assumes that no NAC solution offers post-admission security beyond the occasional re-scan.  Not so, folks, just not so.  Good post-admission security not only provides flexibility with respect to pre-admission policy, but also provides a means to verify what the endpoint just told you was true.  It's not rocket surgery, for heaven's sakes.  OK, it sort of is, but we're really good at it.

Adaptive Enforcement

Should enforcement be through VLAN placement or ACL assignment?  The truth is that as long as you're in the boat of not being able to change or modify the initial decision it doesn't really matter.  Say an endpoint joins the network and needs remediation.  So it gets put in the remediation VLAN (or has a remediation-specific ACL applied to its port), and gets automagically remediated.  Now what?  How does the endpoint get placed into the "you are now a good endpoint" VLAN (or have the "you are now a good endpoint" ACL applied to its port)?  The answer is that, absent not-so-great methods like CLI or SNMP, the mechanism doesn't exist.

I'm picking (mainly) on NAP here, but really all software/infrastructure based NAC solutions assumed managed assets and are therefore in the same boat.  You have to have some mechanism to discover and profile devices currently on your network; you need the backstop of post-admission security; and you have to be able to change/revoke an endpoint's access as its compliance status changes.  We can argue about a whole litany of things (enforcement methods, device management, policy lifecycle, etc); but none of it is really relevant those three basic needs are met.

Microsoft Takes a NAP on Non-Managed Devices

Since the release of NAP compatible 2008 server ostensibly inaugurated this blog, I thought this week would be good time to revisit the big hairy beast that is Microsoft NAP.  While it's true that our (read: my) primary focus around infrastructure-based NAC is biased towards what the IETF ratifies, we remain NAP partners and continue to follow its progression.  One of the more common RSA questions, from press, analysts, and booth visitors was how we (specifically) and NAC pure plays (generally) compete with Microsoft NAP.  The answer, of course, is that we don't.  And the smart ones won't even try.  Here's why:

First, it's worth exploring what NAP is, what it isn't, and where its focus lies.  While some amount of "NAC is stupid/dead/bad" hay has been made over the number of NAC "Standards" offered, the truth is that CNAC, MS NAP, and TNC don't really differ by all that much.  They all look to leverage the initial authentication event to glean endpoint data characteristics; they all have a model for both granting and gating a level of network access based on the combination of endpoint characteristics and user identity; and they all have a fundamental presumption that the endpoint has the capability and willingness to run the endpoint software to make the declarations.  They presume what's connecting to your network are general purpose computing devices, with general purpose operating systems, managed by you, the IT organization.

Now, what's wrong with that model?  The answer, really, is nothing, so far as it goes.  Governing access to your network for your general-purpose computing assets is definitely a problem for you to solve.  MS NAP, with all of it's tradeoffs, is as likely to solve this problem for you as anything else is.  But controlling access for assets you *don't* own, as well for assets that you own but that are specialized in their function (printers, security cameras, cash registers, HVAC controllers, badge readers, the list goes on) is also a problem for you to solve.  Reasonable people can disagree over what percentage this is, and the percentage likely varies by vertical in any event, but that it is "some number greater than zero" is a slam dunk.

This is really where pure play vendors come in, and why (at least here) we openly welcome the advancements of NAP/TNC/NEA.  I've believed for a while now that you will govern your internally-owned desktops and laptops with something other than us.  Symantec.  McAfee. IBM.  Microsoft.  Juniper.  That's a good list; go to it.  But remember the "courage, serenity and wisdom" prayer?  You not only need a solution that brings governance to the class of non-managed (both unmanaged and unmanageable) devices, but you also need one that has the wisdom to know the difference.  The basic visibility of detecting and classifying what's connecting to your network is a tough thing to make a must have or to wrap ROI around (around which to wrap ROI?  Honestly, some times it's just better to end with a preposition).  But the truth is that discovery and classification do have value, and are critical pieces of any kind of reasonable NAC policy (They may be critical for an unreasonable NAC policy, but I try not to worry with such things).

So, what you get with NAP is governance for the Windows Vista and Windows XP SP3 devices that you own and administer.  What you don't get is governance for anything else.  That's a perfectly fair tradeoff, and a perfectly appropriate thing (I think) for Microsoft to go off and solve.  But what's left?  Why might you want some other (granted, integrated) NAC solution to help?  Here's why:

State:  What's on your network?  The importance of answering this basic question can't be overstated.  At least in all the environments I've ever had to manage, it's a moving target.  Minute by minute, second by second.  Which means that robust, real-time state of all devices is the first step.

Classification:  An extension of the above, this is where you get an additional level of detail about the endpoint, as well as when you split between, say "MS NAP" devices and "Non MS NAP" devices.

Post-Connect Monitoring:  Wheels within wheels.  This not only helps make your network safe by providing another layer (importantly, defined within the same management console, under the same policy construct), but it also gives some flexibility on making the front-end admission decision.  Reasonable people continue to disagree over whether organizations will accept a NAC policy that restricts at entry based on what many consider to be IT failures (firewall, patch, AV status), and Microsoft's own internal NAP deployment called out that tension.  A strong monitoring policy post-connect is the best way to cross that chasm.

Integrated Rollups and Status:  Duh

Enforcement:  Duh. See previous post and don't forget the C.

Ticking Away the Moments

(Errata Notice:  When I read the entries on Steve's blog, my eyes interpreted shanna as "Shanna;" and I incorrectly assumed that someone else at Juniper had posted an entry in Steve's blog.  I've corrected the entry below.  My apologies to Steve and my thanks to Chris Radkowski for pointing out my error.)

So, I've been in a discussion with Steve Hanna over on the Juniper NAC blog about the state of TNC NAC standards within the IETF (Steve's initial post that spurred my comment is here; his response post is here).  Network World's Tim Greene also had a newsletter item on the same topic.  Setting aside that Tim's view is necessarily more pessimistic than Steve's, there is no doubt in my mind that progress on the standards front is a good thing for all NAC vendors.  A rising tide that lifts all boats.  The strong-armed quarterback that completes the West Coast Offense.  The Turn card that completes the flush.  You get the idea.

Steve's initial post was around IETF's acceptance of the TNC client-server protocol definitions.  My follow-on question involved the progress, if any, of standards adoption for the enforcement pieces.  Steve's response was both helpful and encouraging, and I look forward to reading additional details as they come.  I'm not looking to minimize the importance of standardizing client-server communication within a NAC framework. However, we continue to impatiently tap our fingers waiting for a similar level of progress on the enforcement front.  The reasons for this are (a) it's "easier," at least politically, since there is broad agreement as to where enforcement should be done; and (b) enforcement puts the C in NAC, since there can be no control without enforcement.  As is obvious, NAC without the C just becomes NA.  Not an acronym (abbreviation, actually, but that's a whole other rant) that any of us wants.

Now, to be sure, every NAC vendor in the space claims "some" kind of enforcement capability, including Mirage.  And it remains more likely than not that at least some of these enforcement mechanisms remain relevant post standards adoption, for unmanaged assets if nothing else.  After all, whether you're talking about NAP, TNC, NEA or some kind of hybrid, the enforcement elements look pretty much the same, which is to say based around a RADIUS authentication handshake, which holds a two-fold assumption that the asset is (a)managed by the organization and (b) capable of running some kind of host-based software (VPN client, 802.1x supplicant).

Still, for that class of managed assets running host-based software, I don't believe there is significant disagreement that RADIUS extensions are the way to go.  RFC 3580 was great, in that it defined a standard way of providing VLAN assignment and port-based filters to connecting users according to the policy defined on the RADIUS server.  The missing piece is the ability to change those attributes post entry outside the context of an authentication handshake.  This is important not only for post-admission security but also for remediation, since remediation by definition changes the status of the endpoint.  By extending the traditional RADIUS Packet of Disconnect to be a more generic Change of Authorization, RFC 3576 fills this gap, seamlessly, across switching platforms, and for all connection methodologies (wired, wireless, VPN).

So, what's the holdup?  I honestly have no idea.  As Steve points out, RFC 3576 is 5 years old now.  5 years.  5.  Five.  Why hasn't the unknown-pantone-blend switching company (Are they gray?  Blue?  Green?  Pick one!) just gone ahead and implemented it?  It would sell more of their switches.  It would keep them a relevant player in the NAC arena irrespective of whatever else happens (see above).  And it's just the right answer.  My hope is that Juniper (now that they're in the switching business), along with maybe Enterasys, will provide the competitive push that is clearly necessary.  Hope is good.  Hope springs eternal.  Time, I suppose will tell us whether Hope fell down and sprained her ankle.  Let's just please make the time something less than another 5 years.