DARPA Thinks the Future of Surveillance Looks Like Siri

Dan Kaufman, director of the Information Innovation Office at the Defense Advanced Projects Research Agency, comes to DARPA from a long tenure in the warp-speed tech industry. When you hear him discuss innovation, you can tell that if he isn’t already impatient with the pace of government, he soon will be. “Innovation doesn’t know a time table,” he said, during a live webcast interview with Defense One on Thursday.

Kaufman has been operating at the intersection of the tech world and government and so his take on the future of privacy, big data and computation is unique. He’s an alumnus of Dreamworks, had a hand in a videogame company and worked as a tech industry lawyer. Hyperbolic warnings about a cyber Pearl Harbor don’t ring right to him. He’s against responding to unnamed “existential threats,” defined as the notion that the human race itself poses a massive danger to U.S. cyber-security. “I push pretty hard against that,” he said.

Kaufman perceives an innovation gap, with the private sector leading in many areas where not long ago, research breakthroughs were mostly the domain of government. One way he’s bridging that gap: The agency’s most recent big data research effort is part of the DARPA Open Catalog, an initiative to open up much more of the software and science research that DARPA sponsors to the public.

At some point in the not-too-distant future, Kaufman believes we will worry about cybersecurity not in the way we do now, which is a constant state of panic, but the way we worry about unlocked doors and holes in fences. Our networks and infrastructure will never be perfectly secure, but we can arrive at a point where they are no more vulnerable than any physical structure.

He also says that we don’t have to make a choice between privacy and security. Kaufman perceives a technological fix, better encryption.

“We feel like there’s a slider that goes back and forth. Where, either we should collect everything, which feels bad, or we should collect nothing. And that also feels bad. What if there was a way to collect the data but encrypt it so that people couldn’t use it in a way that wasn’t approved?” he said.

Kaufman is also looking toward a world where data eavesdropping is harder despite the existence of much more data from many more places. He pointed to a successful demonstration, he claimed, of the agency’s PROCEED program, which sought to harness fully homomorphic encryption, a type of super-encryption for cloud environments that was long thought to be impossible until very recently.

Advanced machine learning will also play a key role in DARPA’s goal of helping the Defense Department more effectively manage and process threats. Here’s how this would look: imagine someone stationed at an intelligence, surveillance and reconnaissance facility looking for visual data of interest. Today, he said, this looks like one person in front of a bank of screens, not really seeing anything. In the future, it looks like a person interacting with an algorithm that becomes smarter about you and what you’re looking for as you interact with it.

The future of reconnaissance looks a lot like Siri. Oh yeah, that was a DARPA program, too.