This Pentagon Project Makes Cyberwar as Easy as Angry Birds

In a dimly-lit conference room at Darpa HQ, Roelker shows me what he showed to the brass. He’s dressed in worn jeans, frayed moccasins, and a black t-shirt with a pistol silkscreened in gray. Maybe he shaved a week ago. We joke about my decision to wear a suit jacket, and he warns me not to expect too much from the demo.

“This isn’t, like, fully functioning,” he says. “This is just to get those big ideas we have in Plan X out for folks to think about and consider. ‘Cause it kind of changes a little bit how people reason about cyberspace, how you visualize it, how you interact and navigate with it.” (Darpa wouldn’t allow me to take pictures of the system during the demonstration, nor would the agency agree to provide screenshots.)

The top right corner of the Touch Table announces the mission: “botnet takedown.” A red star in the network constellation represents the botnet’s command-and-control server — and the target of this mock operation. ”So here’s the node… What kind of weapon package, in a military sense,  do I want to hit that node with?” he asks, and he taps on the red server. Four blue wedges, equally sized, pop up around it. Each wedge has an icon — lightning bolts, radiating disk — and a number.

In most video games, players amass bullets, gold or some other kind of resource that they then expend to help them advance through the adventure. The same principle applies to the numbers affixed to the weapons packages here. “Maybe some technologies were more expensive to develop. Maybe it’s more risky or controversial,” Roelker says. “Maybe we spent $5 million building X, and if we use it, there’s a 50% chance we might lose it.”  The numbers are meaningless for now; they’re just meant to convey that there’s some cost attached to every cyberattack. Roelker picks a weapon package called Sonic Boom, which costs him 10 points.

Then he picks a series of “battle units.” Some be used to download a rootkit onto a target machine or launch denial of service attack. (Although Roelker’s careful not to use those terms with me. “In Plan-X, we’re not doing any research into exploits, keyloggers, or rootkits,” he says. “We’re not building any types of those technologies. In this program.”) Other units might measure how the takedown is going. With every battle unit he picks, optimal routes between the nodes pop up in the screen’s background. The links between the digital constellations’ stars grow more intricate. Roelker picks a node to hop on the network, and adds that to the plan.

He then moves over to the far fight of the screen, where there’s a list of odd names, like “Angry Squirrel” and “Blanket Swarm.” One of the first companies Roelker’s team contacted when they started looking for interface designers was Massive Black, the San Francisco illustration firm that worked on everything from Bioshock to Transformers to a new line of G.I. Joe toys. “One idea Massive Black had was this concept of a playbook,” Roelker explains.

In a network conflict, you might do the same type of things over and over again. To Massive Black, that sounded like something out a video game, “like Madden Football. You might have a running play a passing play, a fake… If we do the same type of activities, is there some way to build a template and then just allow a planner to look through all the different plays they have,” Roelker continues. “In this case, we’ll take Angry Squirrel.” A new set of nodes light up on the screen.

Again, the plays themselves are meaningless — just words, not representing any actual offensive tactics. But the idea of pre-made cyberattack that can be launched with a screen tap? That is anything but meaningless. If Roelker’s idea works out, every time a war plan is made in an interface like this, it’ll compile a custom-made software program. Then it will be error checked, and pronounced ready to deploy. Push-button cyberstrikes. ”Looks like my total progress is 100%,” Roelker says. The attack is planned, and ready to commence.