August 09, 2005 | The Pentagon's Computers Knew, Congressman Says

A Congressman and an unnamed former defense official say that a secret Pentagon intelligence group used data-mining software in 2000 that fingered Mohammed Atta and three other September 11 hijackers as terrorists and considered turning over their information to the FBI, according to New York Times reporter Douglas Jehl's front page bombshell.

The information never made it into the 9/11 commission report and a Pentagon spokesman claims not to know about the program (which does not mean it didn't exist).

I'm skeptical of this report for a number of reasons, including that Weldon's new book about the dangers of Iran relies mostly on information funneled to him by Manucher Ghorbanifar, perhaps best known as a shadowy arms dealer from the Iran-Contra affair and whom the CIA has dismissed as a fabricator.

Mr. Weldon has long been a champion of the kind of data-mining analysis that was the basis for the work of the Able Danger team.

The former intelligence official spoke on the condition of anonymity, saying he did not want to jeopardize political support and the possible financing for future data-mining operations by speaking publicly. He said the team had been established by the Special Operations Command in 1999, under a classified directive issued by Gen. Hugh Shelton, then the chairman of the Joint Chiefs of Staff, to assemble information about Al Qaeda networks around the world.

"Ultimately, Able Danger was going to give decision makers options for taking out Al Qaeda targets," the former defense intelligence official said.

[...]

"We knew these were bad guys, and we wanted to do something about them," the former intelligence official said.

[...]

In the summer of 2000, the military team, known as Able Danger, prepared a chart that included visa photographs of the four men and recommended to the military's Special Operations Command that the information be shared with the Federal Bureau of Investigation, the congressman, Representative Curt Weldon of Pennsylvania, and the former intelligence official said Monday.

The recommendation was rejected and the information was not shared, they said, apparently at least in part because Mr. Atta, and the others were in the United States on valid entry visas. Under American law, United States citizens and green-card holders may not be singled out in intelligence-collection operations by the military or intelligence agencies. That protection does not extend to visa holders, but Mr. Weldon and the former intelligence official said it might have reinforced a sense of discomfort common before Sept. 11 about sharing intelligence information with a law enforcement agency.

Maybe that was the institutional culture at the time and certainly the 9/11 Commission report demonstrated that agencies were reluctant to share information or do so efficiently.

But not passing on the information because of visa issues?

I don't buy it.

I think there may be a really interesting story here, but I think Weldon is getting taken, quite willfully, on a ride.

Update: Phillip Carter of Intel Dump shares my skepticism, but adds that regardless of the story's veracity, the story should spark some further debate.

First, it shows that the U.S. has been actively engaged with data-mining for some time, at least since 2000. Second, notwithstanding the spectacular death of Total Information Awareness some time ago, it appears likely that the U.S. continues to engage in data mining work - both R&D and operational work. Third, I think this story demonstrates the promise of data mining techniques, which fuse the disciplines of artificial intelligence, relational databases, and non-obvious relationship analysis (NORA).

Of course, pitfalls remain for the use of these systems. The U.S. must implement robust control measures to ensure systems like Able Danger don't evolve into an extra-Constitutional menace. Some of those control measures might include the use of ex ante and ex post judicial review for these programs; the appointment of a bipartisan commission to monitor these systems; and a procedural mechanism which would allow citizens to learn of the data held by the government about them (similar to what exists now under FOIA, but better).

Ultimately, however, I think the potential promise outweighs the risk. The government's primitive passenger screening system in place on 9/11 identified 9 of the 19 hijackers, according to the 9/11 Commission report. Other data mining systems have helped big time in the efforts to combat terrorism financing. We need these systems, and we need to find ways to mitigate the legal, policy and political risks so that we can put them into action

.
Link.
A couple of things:
1. It shouldn't be a surprise to anyone that intelligence agencies have been using data mining prior to and after the 2001 terrorist attacks. Data mining techniques are a way of making some sense of a piles of data. I used them to make sense of search logs prior to 9/11 and of course, intelligence agencies were using them to make sense of open source, intelligence and signals data.

2. Carter is right that data mining will be a big issue again soon. There's been several signs of this over the last few months, but I don't feel like giving too much away until I get a good story or two out.

3. Carter skips over what will be the salient points in the coming debate.

Should data mining tools respect the difference between U.S. persons and foreigners?

Should that respect apply only to predictive data mining -- the kind that tries to figure out who is a terrorist before they strike or also to data mining to track down criminals after they've committed a crime?

What kind of data should government data mining programs have access to? Commercial data it buys? Commercial data that it subpoenas? Commercial data it asks for? Your phone records? Your medical history? The list of web sites you visit or people you email? The contents of your email?

If encryption shields personal information in databases (your name becomes X3$FSG, for instance) and the government can only see the uncoded results of data mining after proving to a judge that the data mining results might be important to an investigation, does that change what kind of data mining should be allowed?

These will be the real questions that need to be debated.

Posted by Ryan Singel at August 9, 2005 09:06 AM

Trackback Pings

TrackBack URL for this entry:
http://www.secondaryscreening.net/cgi-bin/mt-tb.cgi/224

Listed below are links to weblogs that reference The Pentagon's Computers Knew, Congressman Says:

» Sins of Omission from UNCoRRELATED
Now that the election is over, its apparently safe for the New York Times to report the news again. Rep. Curt Weldon, who was subject of a "Plame faction" CIA push-back after publishing the claim that bin Laden had been... [Read More]

Tracked on August 9, 2005 05:12 PM