Pentagon: Human-thinking machines are the answer to cyberthreats

“Part of the problem in cyberspace is it’s not as if it’s a stable environment, and if you just catch up you’ll be OK,” said Arati Prabhakar, director of the Defense Advanced Research Projects Agency. “Hiring lots of people, training them, patching and praying … that’s not going to get us the kind of scale we need for the future.”

The only way for cybersecurity to evolve rapidly enough to counter threats, she said, was to create machines that can reason like humans, but do it better. To that end, DARPA is sponsoring a Cyber Grand Challenge next August, in which seven teams will compete to create the best program.

Related Story: http://www.washingtonexaminer.com/article/2572767

“What it aims to do is encourage the development of machines, of computers that can do what today is currently done through human reasoning, which is looking at your network, looking at your systems, looking for penetrations, and looking for vulnerabilities,” Prabhakar explained, and which would be capable of “then writing the patches,” all on their own.

That process currently takes “lots of human beings many months” to complete.

“As it turns out, machines don’t know how to do this yet,” Prabhakar added. “But our premise is, and I think it’s starting to play out, that if machines can start learning these things that are currently human reasoning, that they will learn very rapidly, that they will be able to share, and they will be able to scale in a way that human talent really won’t.”

The seven teams of competitors were selected from 104 initial entrants. The winning team will receive a prize of $2 million, while the second and third-place finishers will receive $1 million and $750,000.

“It’s essentially capture the flag, but a league of their own for machines to play their own capture the flag against each other,” Prabhakar said.

“Who do you think is going to win?” she was asked, to which Prabhakar responded, “A machine.”