Can Data Mining Stop the Killing?

24 July 2012 Would Total Information Awareness have stopped James Eagan Holmes? You perhaps remember the fuss. That program by the Defense Department was curtailed when the Senate voted to revoke funding amid a privacy furor in 2003. The project…

24 July 2012

Would Total Information Awareness have stopped James Eagan Holmes?

You perhaps remember the fuss. That program by the Defense Department was curtailed when the Senate voted to revoke funding amid a privacy furor in 2003. The project had been aimed partly at automatically collecting vast amounts of data and looking for patterns detectable only by computers.

It was originated by Adm. John Poindexter—yes, the same one prosecuted in the Reagan-era Iran-Contra scandal—who said the key to stopping terrorism was “transaction” data. For terrorists to carry out attacks, he explained in a 2002 speech, “their people must engage in transactions and they will leave signatures in this information space.”

Enlarge Image

Associated Press
James Eagan Holmes and defense attorney Tamara Brady in Arapaho County district court, July 23.

The Colorado shooter Mr. Holmes dropped out of school via email. He tried to join a shooting range with phone calls and emails going back and forth. He bought weapons and bomb-making equipment. He placed orders at various websites for a large quantity of ammunition. Aside from privacy considerations, is there anything in principle to stop government computers, assuming they have access to the data, from algorithmically detecting the patterns of a mass shooting in the planning stages?

It helps to go back over the controversy at the time. Supporters argued that Total Information Awareness shouldn’t be frightful to Americans—there would be no monitoring of identified individuals unless a warrant was issued. The system wouldn’t be collecting dossiers of personal information or choosing people to spy on, at least initially. It would be raking impersonally through vast streams of data looking for red flags.

The anguishing thing about mass-shooting incidents is that patterns are indeed present. The person usually has a history of causing alarm in people around. The episodes themselves typically begin with a personal setback—a divorce, a firing, an investment failure, getting kicked out of school. And preparations for mass murder certainly leave “signatures” in the “transaction space.”

In 2008, a Wall Street Journal investigation detailed a partial resurrection of Total Information Awareness under new project names. Then, just a few weeks ago, James Bamford, an authority on the ultra-secret National Security Agency, described in Wired Magazine construction of a new facility near Salt Lake City called the Utah Data Center: “Flowing through its servers and routers and stored in near-bottomless databases will be all forms of communication.”

Mr. Bamford says the project is linked to a new National Security Agency specialty, breaking commercial encryption. A few days after the Wired article, Democrat Hank Johnson grilled NSA Chief Keith Alexander at a House budget hearing:

Rep. Johnson: Does the NSA routinely intercept American citizens’ emails?

Gen. Alexander: No.

Rep. Johnson: Does the NSA intercept Americans’ cell phone conversations?

Gen. Alexander: No.

Rep. Johnson: Google searches?

Gen. Alexander: No.

Rep. Johnson: Text messages?

Gen. Alexander: No.

Rep. Johnson: orders?

Gen. Alexander: No.

Rep. Johnson: Bank records?

Gen. Alexander: No.

Mr. Bamford dismisses these as standard agency demurrers. He maintains the agency uses a definition of “intercept” that means people looking at data, not computers looking at data. But it’s also true that much, and potentially all, data is transmitted in the open, wirelessly, even if it happens to be encrypted. Americans already have been told they have no reasonable expectation of privacy when walking down the street in view of networked security cameras. They certainly can have no expectation that their wireless signals will not bounce off any antenna that happens to be in their path.

The argument over Total Information Awareness was not notable at the time for calm deliberation. There might well be a significant difference between government collecting dossiers of information on individuals and the fully automated sifting of vast data traffic for signs of crimes in progress.

What’s more, nobody bothers to deny that both kinds of monitoring already are happening with data communications “outside” the U.S., whatever that means, given that data packets can take many paths between origin and destination.

Sadly, the political blowback of 10 years ago, and the hush-hush with which data mining has proceeded since then, mean we have not been treated to a public exploration of the critical question: whether one sort of data mining originally envisioned—the kind not keyed on a specific individual—can yield actionable warnings. Psychiatric evaluations of dangerousness, we’re often told, are unhelpful because too many fit the pattern who never engage in violence. Can monitoring masses of transaction data help find the real risks? Or would this also lead (as some experts surmise) to unmanageable numbers of false positives?

After the Aurora theater massacre, it might be fair to ask what kinds of things the NSA has programmed its algorithms to look for. Did it, or could it have, picked up on Mr. Holmes’s activities? And if not, what exactly are we getting for the money we spend on data mining?

Wall Street Journal