Centrally Managed Network Security: Hope or Reality?

By on

Ask any IT professional to predict the future of computer networking and you will get different answers.

Guessing the future of technology has always been a dicey business at best. Planning ahead for the next set of security threats and building security solutions that are not vulnerable to these threats is even more difficult, because building network technology is, on the face of it, a purposeful design process, whereas security threats are typically exploitations of sloppy engineering or cleverly threading together unintended uses for software. Future security threats and vulnerabilities are not something that is easy or even possible to anticipate.

The best hope for a security solution that protects enterprise-wide networks while allowing for centralized management is the emergence of standards. Security standards like DCE, Kerberos, SAML, elements of IPv6 and others continue to be the Holy Grail for which we wait. However, the wait continues after decades with little hope for a standards-based security solution in the immediate term. Why wait for a new standard when an existing standard will do?

Let's take a look at some of the security issues facing the IT professional and see what management problems they have in common.

  • All data, including passwords, flows in the clear over most networks.
  • The insider threat is on the rise.
  • Wireless networking means networking is not restricted to wires any more.

In today's computer architectures, all data flows over the network. If the data stream is watched long enough, some juicy morsels are sure to turn up. Password sniffing, watching the network stream for unencrypted passwords, is the most obvious example. Unfortunately, it is just the tip of the iceberg. Uncontrolled access to all network traffic ensures sensitive data will eventually be compromised. Access to the network stream must be limited to only the data bound for the host.

Another huge security problem that has been mainly ignored, because it is hard to solve, is the insider threat. An insider can abuse their privilege to collect sensitive data for the fun and profit of the individual, with no regard for the organization. The insider threat has always been much greater than the hacker threat, but mild-mannered, trusted employees stealing data is not headline material like a 13-year-old hacking the Pentagon. In reality, the disgruntled employee has more motivation and more access to sensitive data than any outsider. The network must not turn into an enabling technology for the disgruntled employee to compromise data or have access to data beyond that with which they are trusted.

Today, wireless networks are commonplace, sending and receiving data packets within corporate meeting rooms, office bays and yes, even beaches. With the expanded network perimeter, it is even more important that each packet coming in be authenticated to make sure it is coming from a valid user. Additionally, each packet going out must be checked to ensure sensitive data is not being broadcast to allow just anyone to pickup.

One way to solve all three of these problems is to control the data packets going in and out of each host. If each host only looks at the packets bound for it, the password sniffing problem is solved. If users are only allowed to access the hosts they need to do their job, they are limited to only the data with which they are trusted. As result, as much as possible, insider threat protection is gained. Finally, if each packet is checked going in and out of a host, the wireless hub would never see sensitive packets, and outsider packets entering via the wireless hub would be read only by the intended host.

Of course, the solution creates its own problems. If the software controlling the host's access to the network can be modified from the host, then all that's been accomplished is to create a little bit more work for the hacker. The hacker simply disables the security software and then proceeds normally. If the policy can be changed on the host it is protecting, the policy is basically useless. Another problem with controlling every host's access to the network is the management of multiple security policies. Setting up policy on an individual host-by-host basis is an insurmountable task. A better solution is a centrally managed policy approach that scales for large organizations and cannot be tampered with by the individual hosts.

The solution in a nutshell is a distributed packet filtering firewall in front of each host on the network. Why packet filtering? There are many higher-layer protocols that could be selected, but it comes down to a question of standards. All higher-level protocols run over the same packet protocols. By controlling the information flow at the packet level, the information flow for all of the subsequent protocols will be controlled, without the need to adopt a new network packet standard. The solution requires simple packet filtering, as opposed to a stateful inspection approach. The more complicated the packet inspection, the greater the impact on throughput. Authenticating the packet's true origin is much more effective than trying to understand the contents of the packet.

Every host on the network today communicates using a network packet standard protocol that can be utilized to perform centralized management. The current accepted and deployed standard makes centralized management a reality, eliminating the need to wait for a new standard to evolve.

The current situation is very similar to the dawn of the Internet. In the beginning, organizations had to be educated on what a firewall was, and why it was a good idea to protect assets from the Internet. Today, no one even considers connecting an organization to the Internet without a firewall. In ten years firewalls have gone from a concept for researchers to a practical requirement understood by the average Internet user. Similarly, the concept of a distributed firewall on every host is a new concept that has not caught on. Ten years from now, no one will consider connecting a computer to a network without a distributed firewall to protect it.

Dan Thomsen is senior principal research scientist for Secure Computing Corporation (www.securecomputing.com).

Secure Computing Corporation is exhibiting at Infosecurity Europe, Europe's largest and most important information security event. Now in its eighth year, the show features Europe's largest free education program, and over 200 exhibitors at the Grand Hall at Olympia, London, April 29 -May 1, 2003. (See www.infosec.co.uk)

For more information, visit www.securecomputing.com

Copyright © SC Magazine, US edition

Most Read Articles

Log In

|  Forgot your password?