Big Brother Gets a Brain

The Pentagon's Plan for Tracking Everything That Moves

The forensic process could be further enhanced by one of DARPA's analysis programs, like LifeLog or Total Information Awareness. After mining license plate numbers from the footage, investigators could identify the car owners. And then dig into the owners' Web-surfing trails, to see if there were any visits to explosive-making sites. And scan e-mail accounts for virulent language. And plumb credit card receipts for big fertilizer purchases.

To the uninitiated, storing and sharing all this information might seem like insurmountably complex tasks. And according to Strat, the CTS manager, the ability to network surveillance cameras over a wide area is "not right around the corner."

Defense and technology analysts have a different view.

Illustration by Richard Borge
Illustration by Richard Borge


Other Stories in This Week's Voice

"The EPA Opens the Door to Testing Bug Killers on People" by Tennille Tracy

"Could Ashcroft Pass a Pop Quiz on the Constitution?" by Nat Hentoff

"Liberia: Diamonds in the Rough" by James Ridgeway

"(CTS) is pretty creepy. And the creepiest part about it is that it's not all that sophisticated," said Lee Tien, a senior staff attorney with the privacy-rights proponent Electronic Frontier Foundation.

DARPA has mandated that the CTS demonstrations be done only with readily available, "off the shelf" equipment—the kind of stuff you could get at You could find slightly less diesel versions of the gear at

So getting the cameras will be easy. What may be harder is handing off information—a description of a suspicious vehicle, say—from one camera to the next. These lenses will be separated by hundreds, even thousands, of meters. And "appearances can change dramatically" in those distances, Johns Hopkins University senior research scientist Chris Diehl said. Slight variations in light or in the camera's angle can make a car look very different to a mechanical eye. "If you read the literature, there really isn't a proven method" for solving this problem, he said.

Yet this obstacle seems surmountable. In a CTS simulation conducted by software developer Alphatech, a car could be tracked over 10 kilometers with accuracy of 90 percent or better with cameras placed 400 meters apart. The percentage went up, of course, as the cameras moved closer together.

CTS is but one of an array of private and public sector programs to sort through the ever expanding amount of surveillance imagery. University of California at San Diego's Computer Vision and Robotics Research lab just received a $600,000 grant from a Defense Department counterterror group for a CTS-like project. At Los Alamos National Laboratory, Stephen Brumby is using genetic algorithms—programs that are bred from smaller components of code—to automatically analyze satellite pictures.

At the Sarnoff Corporation, a project dubbed Video Flashlight would morph cameras' views into a single three-dimensional model. Using a joystick, a security officer could maneuver through this simulated world as though playing a game of Half Life or Grand Theft Auto.

In order for Video Flashlight to work, however, it would have to use stationary cameras. CTS doesn't have that limitation; it's supposed to function with drones and other battlefield sensors. That's one of the reasons's John Pike thinks the program could have a legitimate military function—"to the extent that it is relevant to urban operations, as opposed to the running of a well-oiled police state."

Combat in cities "tends to quickly degenerate into small firefights," Pike explained. It's a lot harder to know what's happening in a crowded city than it is in an open desert. Radios cut out quicker; drones and satellites have a harder time peering through the concrete canyons and narrow passageways of urban life. CTS could restore some of that sight, giving U.S. generals a "broader situational awareness."

This assumes, of course, that CTS has anything to do with urban combat. If it does, it'd be a surprise to some of the businesses bidding for the CTS contract.

"The primary application is for homeland security," said Tom Lento, a spokesman for the Sarnoff Corporation.

"The whole theme here is homeland security," added Northrop Grumman's De Witte.

Strat disagreed. "DARPA's mission is not to do homeland security," he said.

In a presentation to industry, DARPA noted, "CTS technology will be demonstrated only within the observable boundaries of government installations where video surveillance is expressly permitted, and operational deployment areas outside the United States where it is consistent with all local laws."

But in an interview, Strat did admit that "there's a chance that some of this technology might work its way" into domestic surveillance programs.

In the test at Fort Belvoir this year the aim is to track 90 percent of all of cars within the target area for any given 30-minute period. The paths of 1 million vehicles should be stored and retrievable within three seconds. A year after that, CTS is supposed to move on to testing in an urban combat setting, where it will gather information from 100 mobile sensors, like drone spy planes and "video ropes" containing dozens of tiny cameras.

Shortly thereafter, CTS could be keeping tabs on a city near you.

"This is coming whether we like it or not," said Jim Lewis, with the Center for Strategic and International Studies. "It's not how do we stop the tidal wave. It's how do we manage it."

Noah Shachtman edits the blog

VOICE Question of the Day: If domestic use of CTS technology is inevitable, what assurances would you need from the government to be comfortable with it?

« Previous Page