Two years ago, a pair of Navy information leaders decided to attack their own networks—and not just once or twice a year during scheduled exercises, but far more frequently, and unannounced. Now they’re trying to get the rest of the Navy—and the Pentagon—to follow suit.
Their experiment showed that frequent, automated red-teaming reveals which vulnerabilities are the most dangerous, the easiest for an attacker to exploit with the highest impact—information they wouldn’t have otherwise, said Aaron Weis, the Navy’s chief information officer, or CIO, and Scott Bischoff, the command information officer at the Naval Postgraduate School.
And it’s far more effective than the way the Defense Department currently handles cybersecurity: with checklists of steps taken, patches implemented, and so on.
“It’s a very compliance-driven mentality, like an audit… and it’s wrong,” Weis told Defense One. “Cybersecurity is not a compliance problem.”
Treating cybersecurity like a checklist does answer one question: whether the officer, team, or company charged with “cybersecurity” has done their job to some agreed-upon level of performance—basically, whether they’ve fulfilled the terms of a contract or the parameters of an assignment. It’s an approach that works well for a bureaucracy, but it’s not the best way to actually secure your networks, Weis said.
“We’ve got…15 to 20 years of track record using a compliance mentality that says it doesn’t work, right? Because we continue to be exercised by our adversaries in cyberspace,” he said.
Weis says the Pentagon needs to measure its networks’ suitability for combat the same way it does for soldiers, sailors, tanks, and ships: through the concept of military readiness.
Such an approach would mean prioritizing the biggest problems first, with second-tier or complicated ones set on slower paths to fixing.
“There’s ‘ready to fight tonight.’ But if you are a carrier strike group and you’re deploying in three months, are you on a path to being ready? You manage your readiness on a day-to-day basis and it’s a function of a whole bunch of things,” he said. “Do we have the right people? Are they trained? Are they qualified, or deficient? Do we have the equipment?”
But Weis had to show that getting to a state of “readiness” in cyberspace is a matter of constant testing and drilling, not filling out compliance forms.
He needed a safe space where he could understand readiness without exposing huge problems to adversaries or taking essential naval networks offline. He went to the Naval Postgraduate School, or NPS, in Monterey, California.
NPS’ Bischoff says the school lends itself to experimentation because it’s on the California research and education network, not a Navy network.
“I have a lot of authorities here that other naval units wouldn’t have. It provides the Navy a small chunk of terrain here to do things,” like test new cybersecurity concepts, he said. “We have a somewhat unique terrain here to maybe take a little bit of risk on and try some things. That’s important for a STEM school, you know. We’re very research-intensive, so it’s right up our alley.”
So NPS forged a cooperative R&D agreement with Rebellion, a defense-focused software startup founded by Chris Lynch, the former head of the Defense Department’s Digital Service.
Rebellion brought in a tool called Nova that does automated red-teaming on networks. But it doesn’t just hammer away at the vulnerabilities it finds like it’s running through a checklist.
“It can identify the system, understand its patch level, catalog its vulnerabilities according to what’s generally available, and then try to run an automated exploit against it, based on what it knows,” said Weis.
The process revealed a lot more information than just a list of vulnerabilities to be patched. They learned which vulnerabilities were the most important, the easiest to attack, which ones let the attacker gain wider network access, things they would have learned only if a real attacker was hitting the system and strategizing its next move. That means that different vulnerabilities had a priority that a checklist completely missed.
Said Weis: “Right now, because of our compliance approach, we focus on patching every vulnerability, right?…In no particular order, by the way. We just line them up and knock them down. And what that approach…disregards is the question: Can that vulnerability actually be exploited?”
Some vulnerabilities that seemed large were very difficult to exploit, and some that seemed small were a lot more dire than their position on the checklist would indicate. Running similar experiments on actual DOD networks would probably reveal a similar result: that the Department is not managing its cyber vulnerabilities with a real-world understanding of how an attacker would actually approach the network.
Bischoff said similar red-team experiments involving humans are great, but they happen once or twice a year. “It’s a snapshot, right. I don’t want a snapshot. I want a movie throughout the year.”
Weis said he’s building off of that experiment.
“We’re in the process of…nominating a set of [Navy] programs that are volunteering to go first and to start to use this different approach,” he said. “We’ve been having leadership-level discussions here since kind of before the holidays last year. It started with a one-on-one with the [Chief Naval Officer] and it’s kind of moved up from there.”
Said Lynch: “Aaron [Weis] is trying to completely change what they’re doing,” not only in the Navy, but potentially across the military.
That is going to be important if the Defense Department is going to move toward the highly-networked joint all-domain command and control, or JADC2, vision at the heart of its plans for the next decade. The more networks, computers, drones, satellites, etc.. are all linked together, the more unmanageable a checklist approach becomes. The only solution will be to assume your enormous networked war machine is under attack at its weakest points—because it is.