Trailblazer The Secret Sharer
>While most of the N.S.A. was reeling on September 11th, inside SARC the horror unfolded “almost like an ‘I-told-you-so’ moment,” according to J. Kirk Wiebe, an intelligence analyst who worked there. “We knew we weren’t keeping up.” SARC was led by a crypto-mathematician named Bill Binney, whom Wiebe describes as “one of the best analysts in history.” Binney and a team of some twenty others believed that they had pinpointed the N.S.A.’s biggest problem—data overload—and then solved it. But the agency’s management hadn’t agreed.
Binney, who is six feet three, is a bespectacled sixty-seven-year-old man with wisps of dark hair; he has the quiet, tense air of a preoccupied intellectual. Now retired and suffering gravely from diabetes, which has already claimed his left leg, he agreed recently to speak publicly for the first time about the Drake case. When we met, at a restaurant near N.S.A. headquarters, he leaned crutches against an extra chair. “This is too serious not to talk about,” he said.
Binney expressed terrible remorse over the way some of his algorithms were used after 9/11. ThinThread, the “little program” that he invented to track enemies outside the U.S., “got twisted,” and was used for both foreign and domestic spying: “I should apologize to the American people. It’s violated everyone’s rights. It can be used to eavesdrop on the whole world.” According to Binney, Drake took his side against the N.S.A.’s management and, as a result, became a political target within the agency.
Binney spent most of his career at the agency. In 1997, he became the technical director of the World Geopolitical and Military Analysis Reporting Group, a division of six thousand employees which focusses on analyzing signals intelligence. By the late nineties, the N.S.A. had become overwhelmed by the amount of digital data it was collecting. Binney and his team began developing codes aimed at streamlining the process, allowing the agency to isolate useful intelligence. This was the beginning of ThinThread.
In the late nineties, Binney estimated that there were some two and a half billion phones in the world and one and a half billion I.P. addresses. Approximately twenty terabytes of unique information passed around the world every minute. Binney started assembling a system that could trap and map all of it. “I wanted to graph the world,” Binney said. “People said, ‘You can’t do this—the possibilities are infinite.’ ” But he argued that “at any given point in time the number of atoms in the universe is big, but it’s finite.”
As Binney imagined it, ThinThread would correlate data from financial transactions, travel records, Web searches, G.P.S. equipment, and any other “attributes” that an analyst might find useful in pinpointing “the bad guys.” By 2000, Binney, using fibre optics, had set up a computer network that could chart relationships among people in real time. It also turned the N.S.A.’s data-collection paradigm upside down. Instead of vacuuming up information around the world and then sending it all back to headquarters for analysis, ThinThread processed information as it was collected—discarding useless information on the spot and avoiding the overload problem that plagued centralized systems. Binney says, “The beauty of it is that it was open-ended, so it could keep expanding.”
Pilot tests of ThinThread proved almost too successful, according to a former intelligence expert who analyzed it. “It was nearly perfect,” the official says. “But it processed such a large amount of data that it picked up more Americans than the other systems.” Though ThinThread was intended to intercept foreign communications, it continued documenting signals when a trail crossed into the U.S. This was a big problem: federal law forbade the monitoring of domestic communications without a court warrant. And a warrant couldn’t be issued without probable cause and a known suspect. In order to comply with the law, Binney installed privacy controls and added an “anonymizing feature,” so that all American communications would be encrypted until a warrant was issued. The system would indicate when a pattern looked suspicious enough to justify a warrant.
But this was before 9/11, and the N.S.A.’s lawyers deemed ThinThread too invasive of Americans’ privacy. In addition, concerns were raised about whether the system would function on a huge scale, although preliminary tests had suggested that it would. In the fall of 2000, Hayden decided not to use ThinThread, largely because of his legal advisers’ concerns. Instead, he funded a rival approach, called Trailblazer, and he turned to private defense contractors to build it. Matthew Aid, the author of a heralded 2009 history of the agency, “The Secret Sentry,” says, “The resistance to ThinThread was just standard bureaucratic politics. ThinThread was small, cost-effective, easy to understand, and protected the identity of Americans. But it wasn’t what the higher-ups wanted. They wanted a big machine that could make Martinis, too.”
The N.S.A.’s failure to stop the 9/11 plot infuriated Binney: he believed that ThinThread had been ready to deploy nine months earlier. Working with N.S.A. counterterrorism experts, he had planned to set up his system at sites where foreign terrorism was prevalent, including Afghanistan and Pakistan. “Those bits of conversations they found too late?” Binney said. “That would have never happened. I had it managed in a way that would send out automatic alerts. It would have been, Bang!”
Meanwhile, there was nothing to show for Trailblazer, other than mounting bills. As the system stalled at the level of schematic drawings, top executives kept shuttling between jobs at the agency and jobs with the high-paying contractors. For a time, both Hayden’s deputy director and his chief of signals-intelligence programs worked at SAIC, a company that won several hundred million dollars in Trailblazer contracts. In 2006, Trailblazer was abandoned as a $1.2-billion flop.