Digging through data for omens
The government begins sifting databases to find clues to terrorism in the making
Annoyed at having to remove your shoes at airport security? Just wait. This month, the Transportation Security Administration began testing a screening program on some Delta Air Lines flights that would subject passengers to a whole new level of scrutiny. Here's the plan: Book a seat, and a computer data-sifting process matches your name, address, birth date, and ticket-purchasing information against financial and commercial databases and government watch lists. The goal is to verify your identity and look for any hint of a security risk. "Our system today is not as precise as it ought to be," says TSA spokesman Robert Johnson. "We don't want to waste more time focusing on grandmas."
If government officials have their way, the TSA's test is just the beginning of a new approach to security called data mining. Next year, the TSA's system will be fully in place, coding all travelers by color: green if they trip no alarms, yellow for further screening, or red if the system flags them as too dangerous to allow on board. Other agencies could deploy far more powerful technologies that would screen us all, to pinpoint among hundreds of millions of innocents anyone planning acts of terrorism. The technology would comb the vast amount of information--purchase records, E-mail and phone logs, travel arrangements--that people generate in their daily lives, looking for telltale patterns such as a purchase of fertilizer, a rented truck, and a series of trips to the Middle East.
The quest has had a rocky start. The Department of Defense's Total Information Awareness (TIA) program, a five- year effort, has committed $260 million so far to develop tools for large-scale data mining. But it ran into a public-relations fiasco after Vice Adm. John Poindexter of Iran-contra fame was named to lead the project. Its Orwellian name and a logo (since dropped) showing an all-seeing eye atop a pyramid didn't help. Congress recently voted to require the Pentagon to justify the program and seek approval before monitoring citizens. Yet politics and privacy fears have not halted the quest: TIA research is going strong, and smaller operations like TSA's are getting underway. The true stumbling block may be technology.
It's a delicate dual challenge: accurately spotting suspicious patterns across multiple databases while minimizing false alarms and safeguarding individual privacy. "It's a project similar in scale to putting a man on the moon," says Usama Fayyad, a data-mining specialist who is CEO of digiMine. "It's going to take a national commitment, the best brains in the country, and many years to do it right."
Scientists were among the first to apply mathematical pattern-finding tools to databases--Fayyad led a data-mining project to identify distant celestial objects called quasars amid billions of ordinary stars and galaxies. Companies also mine data on consumers, analyzing buying habits to spot potential customers and monitoring credit card or phone use to detect fraud. Computers watch, for example, for a known credit card fraud pattern: a small gasoline purchase (to see if a card is valid) followed by a much bigger charge. Or they look for a common factor in events that seem unrelated--say, a series of insurance claims in the same ZIP code.
advertisement
