I recently talked to some security friends on a CloudBytes podcast recording that will be coming out in a few weeks. One of the things that came up was the idea of an air gapped system or network that represents the ultimate in security. I had a couple of thoughts that felt like a great topic for a blog post.
The Gap is Wide
I can think of a ton of classical air gapped systems that we’ve seen in the media. Think about Mission: Impossible and the system that holds the NOC list:
Makes sense right? Totally secure unless you have Tom Cruise in your ductwork. It’s about as safe as you can make your data. It’s also about as inconvenient as you can make your data too. Want to protect a file so no one can ever steal it? Make it so no one can ever access it! Works great for data that doesn’t need to be updated regularly or even analyzed at any point. It’s offline for all intents and purposes.
Know what works great as an air gapped system? Root certificate authority servers. Even Microsoft agrees. So secure that you have to dig it out of storage to ever do anything that requires root. Which means you’re never going to have to worry about it getting corrupted or stolen. And you’re going to spend hours booting it up and making sure it’s functional if you ever need to do anything with it other than watch it collect dust.
Air gapping systems feels like the last line of defense to me. This data is so important that it can never be exposed in any way to anything that might ever read it. However, the concept of a gap between systems is more appealing from a security perspective. Because you can create gaps that aren’t physical in nature and accomplish much of the same idea as isolating a machine in a cool vault.
By now you probably know that concepts like zero-trust network architecture are the key to isolating systems on your network. You don’t remove them from being accessed. Instead, you restrict their access levels. You make users prove they need to see the system instead of just creating a generic list. If you consider the ZTNA system as a classification method you understand more how it works. It’s not just that you have clearance to see the system or the data. You also need to prove your need to access it.
Building on these ideas of ZTNA and host isolation are creating virtual air gaps that you can use effectively without the need to physically isolate devices. I’ve seen some crazy stories about IDS sensors that have the transmit pairs of their Ethernet cables cut so they don’t inadvertently trigger a response to an intruder. That’s the kind of crazy preparedness that creates more problems that it solves.
Instead, by creating robust policies that prevent unauthorized users from accessing systems you can ensure that data can be safely kept online as well as making it available to anyone that wants to analyze it or use it in some way. That does mean that you’re going to need to spend some time building a system that doesn’t have inherent flaws.
In order to ensure your new virtual air gap is secure, you need to use policy manual programming. Why? Because a well-defined policy is more restrictive than just randomly opening ports or creating access lists without context. It’s more painful at first because you have to build the right policy before you can implement it. However, a properly built policy can be automated and reused instead of a just pasting lines from a text document into a console.
If you want to be sure your system is totally isolated, start with the basics. No traffic to the system. Then you build out from there. Who needs to talk to it? What user or class of users? How should they access it? What levels of security do you require? Answering each of those questions also gives you the data you need to define and maintain policy. By relying on these concepts you don’t need to spend hours hunting down user logins or specific ACL entries to lock someone out after they leave or after they’ve breached a system. You just need to remove them from the policy group or change the policy to deny them. Much easier than sweeping the rafters for super spies.
The concept of an air gapped system works for spy movies. The execution of one needs a bit more thought. Thankfully we’ve graduated from workstations that need biometric access to a world where we can control the ways that users and groups can access things. By thinking about those systems in new ways and treating them like isolated islands that must be accessed on purpose instead of building an overly permissive infrastructure in the first place you’ll get ahead of a lot of old technical debt-laden thinking and build a more secure enterprise, whether on premises or in the cloud.