How the Air Force Changed Tune on Cybersecurity
Peter Kim, chief information security officer (CISO) for the U.S. Air Force, calls himself Dr. Doom. Lauren Knausenberger, director of cyberspace innovation for the Air Force, is his opposite. Where he sees trouble, she sees opportunity. Where he sees reasons to say no, she seeks ways to change the question.
For Kim, the dialogue they’ve shared since Knausenberger left her job atop a private sector tech consultancy to join the Air Force, has been transformational.
“I have gone into a kind of rehab for cybersecurity pros,” he says. “I’ve had to admit I have a problem: I can’t lock everything down.” He knows. He’s tried.
The two engage constantly, debating and questioning whether decisions and steps designed to protect Air Force systems and data are having their intended effect, they said, sharing a dais during a recent AFCEA cybersecurity event in Crystal City. “Are the things we’re doing actually making us more secure or just generating a lot of paperwork?” asks Knausenberger. “We are trying to turn everything on its head.”
As for Kim, she added, “Pete’s doing really well on his rehab program.”
One way Knausenberger has turned Kim’s head has been her approach to security certification packages for new software. Instead of developing massive cert packages for every program – documentation that’s hundreds of pages thick and unlikely to every be read – she wants the Air Force to certify the processes used to develop software, rather than the programs.
“Why don’t we think about software like meat at the grocery?” she asked. “USDA doesn’t look at every individual piece of meat… Our goal is to certify the factory, not the program.”
Similarly, Knausenberger says the Air Force is trying now to apply similar requirements to acquisition contracts, accepting the idea that since finding software vulnerabilities is inevitable, it’s best to have a plan for fixing them rather than hoping to regulate them out of existence. “So you might start seeing language that says, ‘You need to fix vulnerabilities within 10 days.’ Or perhaps we may have to pay bug bounties,” she says. “We know nothing is going to be perfect and we need to accept that. But we also need to start putting a level of commercial expectation into our programs.”
Combining development, security and operations into an integrated process – DevSecOps, in industry parlance – is the new name of the game, they argue together. The aim: Build security in during development, rather than bolting it on at the end.
The takeaways from the “Hack-the-Air-Force” bug bounty programs run so far, in that every such effort yields new vulnerabilities – and that thousands of pages of certification didn’t prevent them. As computer power becomes less costly and automation gets easier, hackers can be expected to use artificial intelligence to break through security barriers.
Continuous automated testing is the only way to combat their persistent threat, Kim said.
Michael Baker, CISO at systems integrator, General Dynamics Information Technology, agrees. “The best way to find the vulnerabilities – is to continuously monitor your environment and challenge your assumptions, he says. “Hackers already use automated tools and the latest vulnerabilities to exploit systems. We have to beat them to it – finding and patching those vulnerabilities before they can exploit them. Robust and assured endpoint protection, combined with continuous, automated testing to find vulnerabilities and exploits, is the only way to do that.”
I think we ought to get moving on automated security testing and penetration,” Kim added. “The days of RMF [risk management framework] packages are past. They’re dinosaurs. We’ve got to get to a different way of addressing security controls and the RMF process.”