The head of the intelligence community’s advanced research agency is looking at new encryption standards that can withstand future breakthroughs in quantum computing.
Jason Matheny, the director of the Intelligence Advanced Research Projects Activity (IARPA), told reporters Wednesday that within the next 20 years or so, quantum computers could become sophisticated enough to crack some of the most advanced encryption available today.
However, Matheny said, that’s assuming the United States remains the global front-runner in encryption technology. Allowing countries like Russia or China to take the lead, he added, could have major implications for the future of cybersecurity.
“It wouldn’t be good, and it would be good not to be surprised. One of the reasons that we invest is so that we understand where the state of the art is, understand a bit about what the timelines are so that we can deploy quantum-resistant encryption when it’s critical,” Matheny said during a meeting of the Defense Writers Group.
The good news, he added, is that encryption-busting quantum computers won’t likely be available for at least a few decades. IARPA is one of the leading authorities on quantum computing.
“I think we’re still 20-plus years away from a quantum computer that’s relevant to encryption, so we have enough time to think about what forms of encryption we should be deployed that aren’t based on the difficulty of factoring. Still, I think we need to be doing more research on quantum-resistance encryption,” Matheny said.
Despite having a reputation for being “DAPRA for spies,” Matheny said about 85 percent of its work is unclassified. What’s more, he said IAPRA has been partnering with nontraditional companies to maintain its competitive advantage on state-of-the-art technologies.
“We work a lot with industry, particularly small businesses who aren’t the usual suspects in say, defense or intelligence contracts, to ensure that the best and brightest are able to work on our problems,” Matheny said.
Unlike the National Science Foundation or the National Institutes of Health, IARPA runs some of its research competitively through tournaments, where multiple research teams compete against each other toward a common set of technical goals.
By offering cash rewards for solutions to its thorniest technology problems, IARPA has been able to make the most of the innovative and risk-taking culture of U.S. start-up companies — something Matheny said no other country has quite been able to replicate.
“We just offer prize money for anybody who can solve the problem. They don’t have to have a federally approved accounting system, they don’t have to have a contracting office or go through a six-month process to award money. If they can solve our problem, we award them cash,” he said.
What’s more, individuals tapped to develop research through IARPA have gone on to win prestigious awards like the Nobel Prize.
“This attracts people who aren’t used to doing work with the defense or intelligence communities. They really want to do world-class science, and we give them the resources they need to do that,” Matheny said.
For its part, IARPA has dedicated a large part of its research portfolio on machine learning to process raw data, including raw text and raw video.
One of IARPA’s programs, Aladdin, has proven useful in summarizing the content of video footage.
“There is too much data for human eyeballs to look at,” Matheny said.
Several years ago, IARPA reached out to Google to find out more about the advances in video search. The cutting-edge, it turns out, was simply looking through the keywords provided by users uploading the footage.
While that level of video search may work well for, say, YouTube, it doesn’t help the intelligence community parse videos uploaded by terrorist organizations.
“Usually the terrorists are not polite enough to tag their own videos,” Matheny said.
Matheny said IARPA has completed its work on Aladdin and has been transitioning the finished product to the intelligence community.
The agency is also about a year into its work on Deep Intermodal Video Analysis (DIVA), which automatically detects activities within videos.
IARPA looks to use DIVA to process security camera footage to flag when, for example, someone hands off a suspicious package.
That breakthrough could allow security personnel to respond to threats like the 2013 Navy Yard shooting in real-time.
“There was video, there were video cameras, but there weren’t enough human eyeballs looking at those video cameras in order to provide an alarm bell. What we’re doing is developing systems that automatically detect activities within videos,” Matheny said.
Earlier this month, Steven Walker, director of the Defense Advanced Research Projects Agency, told reporters that the U.S. continues to hold the lead in artificial intelligence technology.
While Matheny said he agrees with that assessment, he added that the federal government needs to continue to give appropriate funding to research and development offices.
“Because right now we’re privileged to have a lead doesn’t mean that we’re guaranteed to have that lead forever, and regardless of policy or investment. This was a lead that was earned through it being a priority within science and technology organizations. It’s because federal R&D organizations recognize that machine learning was important, and we allocated funding to it,” Matheny said.