McRaven was prepared to accept me as “a man of high character and integrity,” but he could not get past the original sin of my transaction with Snowden. “He violated the law, so at the end of the day, in divulging that information, you are dealing with a criminal,” he said. “So where is the integrity in that?” (I return to that useful question in chapter 7.) Whatever my efforts to weigh potential harm, whatever consultations I undertook with the government, McRaven believed the things I wrote had consequences that I was not equipped to understand. My incentives, he hinted, might not even incline me to try very hard.
The admiral controlled his voice, but the heat crept back in. “You as a reporter make the call that it’s more important for the public—and I would contend, more important for the reporter—to get that story out before somebody scoops you,” he said. “The nature of your job is to report. That is in your DNA. And I think your default position is to report. And you can always make a case in your own mind why the American people need to know something.”
* * *
—
Long before my first contact with Snowden, the Aspen Institute had asked me to moderate a plenary session. The headliners would be two former directors of national intelligence, Ambassador John Negroponte and Admiral Dennis Blair. Our assigned subject was entitled “Mission Accomplished? Has the Intelligence Community Connected All the Dots?” So it was that six weeks after the first Snowden disclosures, blind luck granted me an hour onstage with two men who had helped define the modern age of surveillance. The format promised substance after more than a month of low-calorie talking points from Washington. My panelists no longer held office, but they had overseen electronic intelligence gathering in its decade of revolutionary growth. Negroponte served as the first DNI, from 2005 to 2007. He helped recast warrantless surveillance programs under President Bush into forms that the FISA Court and Congress, until then kept in the dark, could bless. On Blair’s watch, in 2009 and 2010, some of those programs expanded so fast that the NSA began to choke on the intake.
In a conference call eight days before we met, I invented a “moderator’s privilege” to reframe our topic. The audience would surely expect us to talk about the NSA and Snowden. It was too late to change the printed program, but for our purposes I proposed to rewrite the question to ask, “Is U.S. Intelligence Connecting All the Dots, or Collecting Too Many?”
Blair responded first, annoyed. “I don’t like it much,” he said. “What the hell is there to say about that? Everybody responsible has said it was all properly authorized, supervised. We’re not going to open this program up to the sort of scrutiny that the zealots want.” Shifting pitch to a mocking whine, which I took to be his impression of a zealot, he added, “‘Tell us everything, because we don’t trust you.’”
Negroponte, the career diplomat, tacked toward diplomacy. “I satisfied myself there were proper safeguards in place,” he said in a reassuring baritone. “I more than satisfied myself.”
Blair carried on, still warming up. “The public and the press don’t understand that’s how we do it,” he said. “What’s staggering to me here is there’s no smoking gun that would indicate that this power is being abused. It’s all this, ‘Oh, potential.’” The zealot’s whine returned. “‘Oh, gosh, this is big!’ My experience is that people follow the rules and they have tattooed on their foreheads that you don’t spy on Americans without permission.”
Blair went on for a while. I said Aspen might be the ideal forum to set the story straight. I would ask. They would answer. They would have plenty of allies in the room. Forty minutes later, to their credit, the two men agreed.
On the appointed day, in a ballroom-sized hall atop Aspen’s Roaring Fork Gorge, I opened with a light remark about strange bedfellows. “It’s safe to say that when we set this panel some months ago, my fellow panelists did not foresee they’d be sitting up onstage with a guy who had communicated clandestinely with Edward Snowden.”
An icebreaker, barely worth a chuckle. My kids would have called it a Dad joke. The response did not augur well. Moderator wit, I find, is usually judged on a curve. There is a kind of alliance between speaker and listeners. The audience looks for cues. It wants to respond. This one, unthawed, stared in silence. Neither Blair nor Negroponte cracked a smile.
Right. Straight to it, then. In 2006, the FISA Court had secretly granted the U.S. government the power to collect and store records of every telephone call in the United States. (Collection was never comprehensive in fact, but it was so authorized.) By then the Bush administration had secretly gathered those records for years without notifying the court. “Why on earth would you want all that information?” I asked. “And how does that fit with the boundaries that the American people would expect in terms of privacy?”
Negroponte, on my right, said the Boston Marathon bombing three months back offered answer enough. “One of the reasons you would do it is you have all that data and you detain the Tsarnaev brothers and find out the phone numbers that they’ve been in touch with in Chechnya or wherever,” he said, “and bounce them against these numbers that you have on file, and maybe you’ll find other people who’ve been calling those same numbers in that database.”
“Bounce” had a hand-tossed sound to it. Pull a bad guy’s phone records, give them a shake, and see who bounces out. No one could object to that. Negroponte chose his words carefully. They were accurate as far as they went. You could not say the same for the story line in Washington. Senator Lindsey Graham, among many others, had recently said the NSA searched only the records of people who were “talking to the terrorists.” Negroponte repeated the false claim now. He paraphrased President Bush’s response to news stories about a related surveillance program in January 2006. “It seems like to me that if somebody is talking to al Qaeda, we want to know why,” Bush had said. Negroponte added, “And that’s kind of the underlying philosophy of this program.”
In fact, that was not at all the way the NSA used the call records. The program was designed to find out whether, not why, U.S. callers had some tie to a terrorist conspiracy, and to do so it searched us all. Working through the FBI, the NSA assembled a five-year inventory of phone calls from every account it could touch. Trillions of calls. Nobody needed to plumb that ocean to find the numbers on a bad guy’s telephone bill. What the NSA actually did was called contact chaining, a sophisticated form of analysis that tried to find hidden, indirect relationships in very large data sets. Contact chaining began with a target telephone number, such as Dzhokhar Tsarnaev’s, and progressively widened the lens to ask whom Tsarnaev’s contacts were talking to, and whom those people were talking to, and so on. Sophisticated software tools mapped the call records as “nodes” and “edges” on a grid so large that the human mind, unaided, could not encompass it. Nodes were dots on the map, each representing a telephone number. Edges were lines drawn between the nodes, each representing a call. A related tool called MapReduce condensed the trillions of data points into summary form that a human analyst could grasp.
Network theory called this map a social graph. It modeled the relationships and groups that defined each person’s interaction with the world. The NSA’s analysis touched nearly all Americans because the size of the graph grew exponentially as contact chaining progressed. The whole point of chaining was to push outward from a target’s immediate contacts to the contacts of contacts, then contacts of contacts of contacts. Each step in that process was called a hop.
Double a penny once a day and you reach a million dollars in less than a month. That is what exponential growth looks like with a base of two. As contact chaining steps through its hops, the social graph grows much faster. If the average person calls or is called by ten other people a year, then each hop produces a tenfold increase in the population of the NSA’s contact map. Most of us talk on the phone with a lot more than ten. Whatever that number, dozens or hundreds, you multiply it by itself to measure the growt
h at each hop.
The NSA’s deputy director, John C. Inglis, had testified in Congress just the day before Negroponte and Blair joined me onstage. Inglis said NSA analysts typically “go out two or three hops” when they chain through the call database. For context, data scientists estimated decades ago that it would take no more than six hops to trace a path between any two people on Earth. Their finding made its way into popular culture in Six Degrees of Separation, the play and subsequent film by John Guare. Three students at Albright College refashioned the film as a parlor game, “Six Degrees of Kevin Bacon.” The game then inspired a website, The Oracle of Bacon, that calculates the shortest path from the Footloose star to any of his Hollywood peers. The site is still live as I write this, and it makes for an entertaining guide on hops and where they can take you.
Bacon shared screen credits with a long list of actors. Those were his direct links, one hop from Bacon himself. Actors who never worked alongside him, but appeared in a film with someone who had, were two hops away from Bacon. Scarlett Johansson never worked with Bacon, but each of them had starred alongside Mickey Rourke: Bacon in Diner, Johannson in Iron Man 2. Two hops, through Rourke, connected them. Bacon had no role in the Star Wars franchise, but he could trace two-hop paths to Harrison Ford, Carrie Fisher, Mark Hamill, and James Earl Jones. If you kept on playing you discovered that Bacon was seldom more than two hops away from any actor, however removed in time and movie style. Two hops reached silent film star Charlie Chaplin, Groucho Marx, and Fred Astaire, none of them born the same century as Bacon. Hedy Lamarr? Humphrey Bogart? Two hops. (Each made a film with Eddie Albert.) In a single-industry town like Hollywood, links like these might make intuitive sense. More surprising, if you did not spend much time around logarithms, was the distance traveled by one or two hops through the vastly larger NSA data set. Academic research suggested that a mean of three hops—three links in the chain, the same number Inglis mentioned—could trace a path between any two Americans.
I wanted to take the conversation there. Blair offered an opening.
“Here’s a fact,” Blair said. “The number of times those records were accessed in 2012. Take a guess. Okay, I’m going to give you a multiple choice, Bart. Ten, two hundred fifty, ten thousand, five million?”
He probably knew I could answer that. Inglis had mentioned the figure the day before. This was a setup for the audience.
“Under three hundred, according to the administration,” I replied.
Blair made a gesture of triumph. Point and match. The FBI and NSA had oceans of information about us. Fine. But they hardly ever looked. Who could object to so shallow a dip into waters so deep?
“Let’s talk about what that means,” I said. “We just heard from Chris Inglis yesterday . . . that contact chaining on those numbers, when you pull those numbers, is done to two or three hops. Let’s suppose that the median number of unique contacts for people making phone calls is one hundred over the course of the year.”
I worked through the arithmetic aloud. Multiply by one hundred at every hop. With phone records from three hundred people, the first hop searched the calls of thirty thousand. The second pushed the boundary to three million. On the third hop, contact chaining would map “a potential universe of about three hundred million, which happens to be approximately the population of the United States.”
Dennis Blair, meet Scarlett Johansson. Odds are you know at least a friend of a friend of a friend. She is two hops from Bacon and no more than three from nearly anyone between Hawaii and Maine.
Negroponte leaned in. “Let me interrupt you,” he said. “Are we having a hypothetical discussion here or a real one?”
“It’s just math,” I replied.
“It’s just math,” Negroponte said, dismissing the point. “It’s not what’s actually happened.” Maybe bulk collection of telephone records empowered the NSA to map the communications of anyone in this room, anyone in America. But the people who possessed that power used it with discipline and restraint. “They check themselves every step of the way,” Blair said, “and they are not rummaging around in trillions of records to try to see if they can find something interesting.”
They check themselves. There were supervisors and compliance officers, an inspector general, a general counsel, and a director of national intelligence who made classified certifications that the NSA followed its rules. (The rules were also classified.) Should we trust them as much as they trusted themselves? Should we trust not only Blair and Negroponte, not only President Obama, but also every successive heir to the surveillance machine? Would Blair himself have bequeathed these powers comfortably, had he known back then that Donald Trump would take the wheel in 2017? By then many of his retired peers would sign “never Trump” letters, declaring him unfit for office on grounds that included a reckless attitude toward power.
The answer might depend, in part, on the stakes. How much could someone really learn from a simple list of our calls? The official government position in court was that there is no privacy interest in this kind of “metadata,” a term that means information about information. It is the who/when/where of a conversation, without the words themselves. At Princeton, where I had a visiting fellowship, I had been working through the meaning of metadata with Ed Felten, a computer scientist who served for a time as deputy chief technology officer of the United States. Our conversation kept returning to “embedded patterns” in any voluminous data set. That summer he spelled out what he meant in an affidavit for the plaintiffs in a lawsuit against the NSA. Information science had learned how to pull intimate secrets from very large collections of very small clues. “Individual pieces of data that previously carried less potential to expose private information may now, in the aggregate, reveal sensitive details about our everyday lives—details that we had no intent or expectation of sharing,” he wrote.
The call data records might one day identify unknown terrorists, but that turned out to be a hard problem. A presidential review group concluded later that year that the NSA had not yet made a breakthrough that way. There were other things, very private things, that were easier to discover. With trivial effort the government could identify people who called whistleblowing, drug abuse, rape, or suicide hotlines. Felten did not mention reporters, but journalistic sources were easy to pick out if they did not take uncommon precautions. With access to the call records, Big Data methods could extract the “membership, donors, political supporters, [and] confidential sources” of human rights or protest groups. Cash donations sent by text message, an increasingly popular channel, identified contributors to political parties and religious institutions. Data mining could reliably pick out sexual orientation. It could track the telephonic fingerprints of secret love affairs as they blossomed, peaked, and died. It could distinguish bosses from employees, in part because bosses get their calls returned faster and have fewer qualms about phoning subordinates at night.
When you factored in time and sequence, the results were startling. “A likely storyline emerges,” Felten wrote, when “a young woman calls her gynecologist; then immediately calls her mother; then a man who, during the past few months, she had repeatedly spoken to on the telephone after 11 pm; followed by a call to a family planning center that also offers abortions.” The government may seldom care, may never abuse that knowledge in a given year. But now, for the first time in history, it had acquired the power to do so.
Stewart Baker, a former general counsel of the NSA, leaped early into the fray against Snowden in television appearances, newspaper interviews, and blog posts. He sharply criticized some of my stories, too. But he minced no words about the power of the social graph. “Metadata absolutely tells you everything about somebody’s life,” he said. For purposes of signals intelligence, “if you have enough metadata, you don’t really need content.” Michael V. Hayden concurred bluntly the following spring. “We kill people based on metadata,” he said. “But that’s not wh
at we do with this metadata.”
In Washington, the day before our panel, Representative Bob Goodlatte, a Virginia Republican, asked Robert Litt, the DNI general counsel, whether intelligence officials had really believed the bulk phone collection “could be indefinitely kept secret from the American people?” Litt replied in a rueful voice, “Well, we tried.”
Blair and Negroponte both had second thoughts about that. “Those of us who were senior officials in the intelligence community and so on should have done a much better job of explaining the general principles of these programs, without going into titillating individual cases which do nothing but help our adversaries,” Blair said. “And it’s a kind of a ‘pay me now or pay me later,’ when . . . you’re operating from a defensive crouch of the Snowden revelations and saying, ‘Yeah, but we’re okay. Trust us.’”
* * *
—
Trust became the core of our conversation. On your watch and since, I told both men, policymakers and intelligence leaders routinely deceived the public with assurances that turned everyday language upside down. NSA director Keith Alexander had said flatly in 2012 that “we don’t hold data on U.S. citizens.” I thought that qualified as a naked lie. If Alexander thought otherwise, he must have relied on some secret definition of “hold” or “data” that no listener could be expected to guess. Three months before the first Snowden story, Senator Ron Wyden asked Director of National Intelligence James Clapper in a public hearing, “Does the NSA collect any type of data at all on millions or hundreds of millions of Americans?” Clapper stared down at the table, stroked his shaved head, and balled his hand in a fist. He wore the face of a man who had to eat something unpleasant. “No, sir,” he told the table. He raised his eyes to Wyden and shook his head. “Not wittingly,” he added. Critics, Republicans among them, later accused him of perjury.
Dark Mirror Page 17