As the Defense Advanced Research Projects Agency (DARPA) continues exploring emerging technologies for the Department of Defense, it’s considering the implementation of artificial intelligence (AI) to tackle electric grid cybersecurity and get ahead of 5G deployment.
William Scherlis, director of the Information Innovation Office (I2O) at DARPA, elaborated on two current projects at the March 24 FCW AI Workshop.
“In the area of cyber operations, we have a program RADICS [Rapid Attack Detection, Isolation and Characterization Systems], which is designed to help recover critical portions of the power grid in the event of a full blackout caused by malware,” Scherlis explained.
The program partners with power companies, the National Guard, the Department of Homeland Security, and the Department of Energy to enable a “black start” recovery, a restart to the electric grid without external power, in the event of a cyberattack. Scherlis mentioned the program in context of AI, but the program is generally building new tech to accelerate recovery through improved situational awareness, network isolation, and the ability to adapt to changing cyber situations.
“The idea of this program is how can we understand what is the state of affairs in the grid and incrementally restore service, purge the malware, and bring service back to the most critical assets and then stage that out beyond. They’ve done a number of field trials,” Scherlis added.
I2O is also exploring tech for creating a more resilient, adaptable, and secure 5G system. The Open, Programmable, Secure-5G (OPS-5G) program aims to get ahead of the 5G curve by developing an open source, “plug-and-play” approach to 5G.
“A solicitation has been posted to open up the 5G ecosystem, which is now vertically integrated and closed, generally speaking, and the purpose of this is to improve security to facilitate customization and support of national security needs,” Scherlis said. “Of course, also, by opening it up, it lowers the entry barriers for firms involved both in the hardware and the software.”
While Scherlis did not explain how AI could specifically be applied to these programs, perhaps because it’s still being explored, he did say DARPA’s approach to AI must be considerate of the private sector.
“We’ve been doing this [AI] for more than 20 years already. We know that it requires great architects, we know that it requires good systems planners,” he said. “The challenge in our space is basically how do we do that at arm’s length in the environment of the defense industrial base.”
Further, as with many emerging technologies, AI is always changing. The problems associated with it take new shape as the tech develops and Scherlis acknowledged that.
“As soon as AI successfully takes on a problem, it graduates out of the AI label and it’s no longer considered AI. A corollary of that is that AI problems are always super hard, because once they stop being hard, they’re no longer AI problems,” Scherlis said.