“Data Mining Doesn’t Catch Terrorists”

by on December 11, 2006 · 2 comments

That’s the quickest summary of a paper the Cato Institute issued today, which I co-wrote with Jeff Jonas, distinguished engineer and chief scientist with IBM’s Entity Analytic Solutions Group.

Data mining is the effort to gain knowledge from patterns in data. A retailer can use data mining to sift through past customer interactions and learn more about potential new customers, but it can’t figure out which customers will actually come into a new store. Terrorism is so rare in society that there are no patterns to search for. Data mining has no capability to ferret out terrorists.

It appears that the Automated Targeting System, which made news last week (because of its previously unknown focus on American travelers), uses data mining. It sifts through information about border crossers to assign them a “risk score.”

In a National Journal article published last week, Secretary of Homeland Secretary Michael Chertoff discussed ATS, revealing the need for government officials to get more clear about what they are doing, what works, and what doesn’t work. According to NJ, Chertoff called ATS “the process by which we collect that information and analyze it to see what are the patterns and the relationships that tell us, for example, that a particular telephone number is associated with a terrorist, or something of that sort.”

Comparing the number of a traveler to phone numbers of terrorists is data matching and it is not what ATS does – or at least not the interesting part of what ATS does. Data matching, link analysis, or “pulling strings” is a proven investigative method and, as we discuss in our paper, it’s what could have prevented the attacks of 9/11.

There should be forthright public discussion about whether a program like ATS, or any data mining program, can catch terrorists. Such a program might help turn up ordinary crime, about which there may be suitable patterns to discover. Whether the public would countenance mass surveillance for ordinary crime control is a different question than whether it would accept such methods to prevent terrorism.

Comments on this entry are closed.

Previous post:

Next post: