Many companies have a red team, or several, and they generally share the same purpose—to play the role of an attacker, probing releases new and old for vulnerabilities, hoping to catch bugs before the bad guys do. Few of them, though, focus on a target as ubiquitous as Windows, an operating system that still boasts nearly 90 percent market share for laptop and desktop computers worldwide. When Windows breaks, the whole world hears the shatter.
Putting It Together
The Windows red team didn’t exist four years ago. That’s around the time that David Weston, who currently leads the crew as principal security group manager for Windows, made his pitch for Microsoft to rethink how it handled the security of its marquee product.
“Most of our hardening of the Windows operating system in previous generations was: Wait for a big attack to happen, or wait for someone to tell us about a new technique, and then spend some time trying to fix that,” Weston says. “Obviously that’s not ideal when the stakes are very high.”
Weston wanted to go beyond Microsoft’s historical mode of using bug bounties and community relationships to formulate a defense. He was tired of the reactive crouch, of responding to known issues rather than discovering new ones. He wanted to play some offense.
Drawing inspiration from his experience with whitehat hackers at events like Pwn2Own—and tired of waiting until the competition ended to glean valuable insights into Windows vulnerabilities—Weston began putting together a team that would essentially conduct a Windows-focused hacking contest every day of the year.
Today, members of that team include Jordan Rabet, whom David noticed after Rabet showed off an impressive Nintendo 3DS jailbreak in a 2014 YouTube video. Rabet currently focuses on browser security but also played a key role in Microsoft’s response to the Spectre vulnerability that rocked the computer industry less than a year ago.
Viktor Brange, who lives in Sweden, helped respond to leaked NSA Windows-hacking tool Eternal Blue by sifting through Microsoft code base, ascertaining the severity of various issues to triage. Adam Zabrocki’s deep Linux experience helps tackle kernel and virtualization issues. Jasika Bawa helps transform the team’s findings into actual product improvements. And two other members of the team WIRED spoke with for this story do sensitive enough work that they requested anonymity.
Together, the red teamers spend their days attacking Windows. Every year, they develop a zero-day exploit to test their defensive blue-team counterparts. And when emergencies like Spectre or EternalBlue happen, they’re among the first to get the call.
Again, red teams aren’t novel; companies that can afford them—and that are aware they could be targeted—tend to use them. If anything, it may come as a surprise that Microsoft hadn’t sicced one on Windows until so recently. Microsoft as a company already had several other red teams in place by the time Weston built one for Windows, though those focused more on operational issues like unpatched machines.
“Windows is still the central repository of malware and exploits. Practically, there’s so much business done around the world on Windows. The attacker mentality is to get the biggest return on investment in what you develop in terms of code and exploits,” says Aaron Lint, who regularly works with red teams in his role as chief scientist at application protection provider Arxan. “Windows is the obvious target.”
Training that mindset internally on Windows has already paid dividends. In addition to helping mitigate Spectre and EternalBlue—the team can only say so much about what, exactly, they did in either case—they’ve notched some important wins that helped not only Microsoft, but the entire industry.
At the top of Weston’s list is shutting down a phishing attack used by notorious Russian hacking group Fancy Bear, which Microsoft calls Strontium, by shoring up Win32k, a Windows kernel-driver and popular hacker punching bag.
“In most browser attacks, you first need to compromise what’s called the browser sandbox, and then you need a way out of that sandbox to do what attackers want to do, information theft or persistent access to the machine,” Weston says. “It turns out that this very old and large kernel surface is the ideal place to do that.”
By attacking that surface through the eyes of an adversary, the team found previously undisclosed techniques to leverage it in an attack. Which meant, in turn, that Microsoft was able to ship an update that blocked those same efforts in Windows 10 Anniversary Edition in the fall of 2016. The Windows 10 Creators Update, released six months later, took even further steps to detect kernel exploits.
It’s an important win, and one that may not have come so quickly had Microsoft relied on more traditional methods of vulnerability-spotting.
“What it tends to be is finding the issues that are a little bit beyond the pale in terms of security vulnerability, that might not be a immediately apparent or directly searchable, findable from vulnerability scanning techniques,” Arxan’s Lint says. After all, you can only scan for problems you already know about. A red team finds the ones you don’t.
Running Out the Clock
The members of the red team don’t have a specific quota; they’ll prioritize targets based on things like what they’ve seen hackers exploit in the wild or which features are relatively untested and sensitive.
“We want to emulate the kinds of things we’ve seen in the wild and then take it to the next level,” says Rabet. “People were doing something a couple of years ago; where are they going to go next? And we try to go in that direction.”
At the same time, the team needs to be selective. “Bugs will always be there,” Zabrocki says. “We can’t fix all the bugs in the world,” especially with as big and complex and constantly evolving a product as Windows. Better, then, to focus on broader solutions like kernel anomaly detection, which can help prevent a whole host of woes.
And solving a problem entirely sometimes isn’t even the objective. Every time the Windows red team starts a project, they also start a clock.
“The goal of the timer is to give us an objective cost analysis of what it takes to hack something,” Weston says. “A start-to-finish, median cost to attack something puts an economic tag on a compromise that’s something we can drive up over time, which we think is a good objective metric.” The more time and money a hack costs to execute, in other words, the less likely an attacker will be to pursue it. Weston hands out computer-shaped trophies for particularly good finds.
The red team doesn’t issue patches, of course, which can lead to some frustrations if they find what they view as a pressing vulnerability that ends up not getting a timely fix. “A lot of it depends on the internal mechanisms within the company. It’s a big company. There are a lot of people who want to have a say in how we do things,” says one anonymous team member, who laments that Microsoft can sometimes take months to fix what both internal and external security researchers see as serious issues.
Helping set those priorities is Bawa, who uses the red team’s activity as an “internal barometer” of how effective Microsoft’s endpoint detection products are—especially against attacks they’ve never seen before. “It really comes down to being able to look at their activity as a blueprint for what we might expect from state of the art activity coming from outside of Microsoft.”
Windows will always be a popular hacker target, and Weston’s team is just one piece of Microsoft’s efforts to protect it. But given the sophistication of hackers, whether they’re nation states or criminal syndicates, it’s at least comforting to know that there’s a team in Redmond keeping pace with the bad guys—and even staying one step ahead.