Datamining: The New Weapon In The War On Terrorism?
The data-mining technology needed to support a massive government initiative to ferret out terrorists through analysis of phone records will be costly and computationally intensive, and could compromise the privacy of ordinary U.S. citizens. While it is uncertain if the government is actually using data-mining techniques to sift through the tens of millions of records it has collected from Verizon, BellSouth, and AT&T, it would need supercomputers comparable to IBM’s Blue Gene to derive meaningful information from a dataset so large, says Nathan Hoskin of Planning Systems.
Hoskin estimates that such a system would cost between $20 million and $50 million. To effectively mine the data, the system would use clustering algorithms to focus on relationships between similar data, link analysis to find connections between disparate data, and rule mining to find patterns within the data.
Privacy advocates warn that giving the government unfettered access to citizens’ phone records, even in the name of fighting terrorism, could lead to a host of civil rights violations without ever producing a lead. Critics have compared the possible data-mining initiative to the aborted Total Information Awareness program envisioned by the Defense Department to preemptively combat terrorist attacks by analyzing patterns within a huge repository of electronic data.
Data-mining experts say that even if the phone companies are not turning over customers’ personal identifying information such as names and street addresses, the government could easily retrieve that information from other databases and services.
While data mining does not go as far as wiretapping, privacy advocates warn that the threat is very real. “Listening to the content of calls is more intrusive, but nobody should underestimate the privacy invasion that’s involved in tracing who’s talking to whom,” said the ACLU’s Jay Stanley, adding that mining records of phone calls for terrorists is inefficient and tantamount to labeling the entire U.S. population as suspects.
Abstracted by the National Law Enforcement and Corrections Technology Center(NLECTC) from Federal Computer Week (05/29/06) Vol. 20, No. 17, P. 38; Sternstein, Aliya .