Data Mining Doesn’t Catch Terrorists

WASHINGTON -- In the wake of 9/11, a jittery Washington embraced any measure intelligence and law-enforcement agencies said would help prevent further terrorist attacks. Data mining was dubbed an essential tool in the war on terror, with the agencies arguing that comprehensive monitoring of personal data would assist in catching terrorists. In "Effective Counter-Terrorism and the Limited Role of Predictive Data Mining," Jim Harper, director of information policy studies at the Cato Institute, and Jeff Jonas, distinguished engineer and chief scientist with IBM's Entity Analytic Solutions Group, demonstrate that data mining is costly, ineffective, and a violation of fundamental liberty when used to find terrorists..

Data mining relies on "pattern-based analysis," where the government analyzes private data from large numbers of people to find terrorists. The data includes travel and transactional history. Predictive data mining not only invades citizens' privacy, but registers false positives over 90 percent of the time. The technology needed to obtain more precise results simply does not exist -- a fact that renders data mining useless and potentially harmful when used for this purpose.

The "statistical likelihood of false positives is so high that predictive data mining will inevitably waste resources and threaten civil liberties," argue the study's authors.

"Data mining is not an effective way to discover incipient terrorism," Harper and Jonas write. "It is not well suited to the terrorist discovery problem." In fact, the authors conclude that the information about the key members of the 9/11 plot was available to the U.S. government prior to the attacks. It was "quite feasible" to find and connect the culprits "using the legal authority and investigative systems that existed before the attacks."

Harper and Jonas add: "Better interagency information sharing, investigatory legwork, in pursuit of genuine leads, and better training are what the 9/11 story most clearly calls for," but data mining distracts from those goals.

In order to keep America safe, Harper and Jonas recommend abandoning data mining and putting resources where they matter. "The most efficient, effective approach -- and the one that protects civil liberties -- is the one suggested by 9/11: pulling the strings that connect bad guys to other plotters."

Policy Analysis Paper no. 584: https://www.cato.org/pub_display.php?pub_id=6784