Network Admission Control
"Locks keep honest people honest," or so my father told me when I was young. I'd guess other fathers around the country provided similar sage advice to children trying to understand a nuanced world. But is there a place for security focused on the good guys in networking? There is, and one of the latest additions to this area of security is called Network Admission Control (NAC).
NAC is built around the notion that by asking specific questions of an organization's end hosts you can improve the overall security of a network by improving the compliance of end systems to a given admission policy.
But first some history. Those of you with some experience in managing networks back in the days of IPX and Novell Netware remember a little thing called the "login script." It was useful to map drives in DOS, set various host parameters, and force a user to patch his system when something new came out such as an antivirus update. At the time, users almost always logged in to the network because file and print services were nearly all you could do with the network. Specialized applications were still run through terminal emulation to mainframes or minis, and nobody could figure out a good reason to connect to the Internet yet. So if you didnt log in, you didnt use the network, plain and simple.
Fast-forward to today...Ive been using computers at employers with Windows-based networking for file and print services for close to 10 years but I can't remember actually logging in on start-up, nor can I remember running a login script for probably five years. Even if I had, the Windows file servers are only one of many things I can do using IP within these organizations. Whether accessing internal websites, applications, or the public Internet, none of these functions requires an overall network login. Even if they did, login scripts typically only run at start-up. With many offices using laptops and wireless access, sleep mode or suspending the systems is common rather than fully shutting down. The laptop Im writing this article on hasnt been "turned off" for well over a week.
This is challenging because compliance matters today more than it did in the old days. Virus and worm attacksand increasingly spywareare more common now than ever and with the mobility of devices, the chance of infection is high as devices move to public networks and then back onto the corporate network. And because of the automated nature these attacks propagate through, the legitimate user represents as big a threat to the network as the bad guy who initiated the attack in the first place. This is because without these legitimate systems to aid in the propagation of the attack, the impact to the overall network is significantly reduced.
Beyond compliance, there is the more fundamental question today of basic network admission. The advent of laptops means outside users are increasingly attempting to connect to an organizations resources. Be they contractors or vendors, the ability to differentiate between insiders and outsiders is key to enforcing basic security policies within your organization. Relying on application authentication as the only means of this differentiation is problematic because not all important systems have such controls.
The principle cause of all these problems, and what NAC is attempting to solve, is that tying endpoint compliance or admission to one of many types of network use is challenging because you cant force the user to initiate the compliance process before potentially spreading malware. NAC solves this by moving the compliance event to initial network access. Before you can do any other function on the network, the device must be authenticated and its compliance level, or posture, must be validated.
NAC works like this: Endpoints are configured to run with an agent or without an agent with some corresponding loss of compliance accuracy. These agents respond to queries initiated by the network to identify various attributes of their posture. This can include OS and hot-fix version number, the presence and configuration of a personal firewall, and the .DAT file version running on a systems AV software, as well as when it last ran a full scan. When an endpoint runs without an agent, the system is audited by the network to determine the relative risk level of allowing it on the network.
In either approach, once the network determines the system is authorized and clean, it can be allowed on the network as usual. Systems that fail some element of the authorization or compliance checks can be relegated to quarantined access temporarily, or permanent restricted access (as in the case of a contractor machine). From quarantined access, the entitys network requests can be redirected to a remediation server where the end station receives instructions to bring itself into compliance. Once brought into compliance, the device can be provided proper access.
Choosing how to deploy NAC first depends on the overall architecture you wish to use. Some organizations are more interested in a turnkey approach that allows NAC to be overlaid onto an existing network infrastructure. Others want NAC more tightly integrated into the network fabric itself. In either approach, there are several components to any NAC solution.
- Network Enforcement Device—The network device has the responsibility to initiate the queries to the end station and then relay the responses to the policy server. Today NAC is supported in Cisco routers running 12.3(8)T or later as well as the VPN 3000 concentrator running software 4.7 or later. In late summer 2005, this capability will be extended to LAN switches and, following that, WLAN access points. For the turnkey approach, Cisco Clean Access (CCA) appliances can be added to your existing network in either inline our out-of-band mode to perform the end system checks.
- Agent—For both the turnkey and fabric-integrated approach, agent software is available to respond to the queries from the enforcement device. In the turnkey model this agent collects all the information and provides it to the network. In the fabric-integrated model, the agent communicates with Cisco and third-party software to learn posture information by acting like a broker on the host. Once collated, this information is then presented to the network.
- Policy Server—The policy server acts as the repository of the posture information required for clean connectivity into the organization. Depending on which deployment model you choose, the policy server either defines compliance policy locally or communicates with third-party policy servers to assess specific credentials. In either case, the policy server makes the final admission decision, which it sends to the network device.
For organizations considering NAC deployment, an important point to remember is that this is relatively new technology. Like any new technology, early adopters will suffer bumps or bruises during deployment. The size of your network can be a big factor here. NAC is already successfully deployed in the small and medium-sized business markets, as well as specific pockets of larger enterprises. In these environments the variables introduced by NAC can be contained and a smoother rollout can be achieved. Larger organizations will want to take more care in evaluating how NAC fits in with its overall network authentication and endpoint security strategy.
While NAC won't stop a determined user bent on introducing viruses and worms into your network, looking back at my father's advice from the beginning of this article: NAC doesn't need to. If NAC is only successful in preventing the systems that arent trying to be active conduits of malicious traffic from attacking the network and keeping unauthorized systems off the network, it will have already performed a valuable service for an organization's IT infrastructure. NAC surpasses most other recent security innovations in importance and utility. Organizations hoping to enforce endpoint admission and compliance policies should give NAC careful attention and begin considering how such a technology would work in their environment.