By Jared Chausow
By Katie Toth
By Elizabeth Flock
By Albert Samaha
By Anna Merlan
By Jon Campbell
By Jon Campbell
By Albert Samaha
Ponemon is more concerned about process than the data itself. "Total privacy does shelter bad guys, there's no question about that. But transparency is also good," he argues. "There should be some labeling or notice." In theory, consumers and investors could punish offending companies by channeling their money elsewhere. Without honest managers, though, the free market's self-correcting mechanism never gets a chance to kick in.
Librarians have filled their listservs with e-mails sharing strategies for resisting law enforcement attempts to grab hold of their users' book lists. But the corporate world doesn't foster that kind of purist culture. When the Federal Bureau of Investigation came knocking for the names of scuba divers this spring, the Professional Association of Diving Instructors forked over a roll of more than 2 million certified divers without so much as being served a subpoena.
The feds were acting on no specific threat, just a hunch that someone might attack that way. And again, these data dumps are just attempts to do good. Would Attorney General John Ashcroft's new TIPS campaignthe Terrorism Information and Prevention Systemencourage people like the mole at the grocery store chain to spill info into the tanks of unethical investigators?
The Department of Justice, which seeks informants in utility, cable, and other such industries operating in communities, denies that it will cultivate sources placed in data-mining operations. "This makes TIPS sound so much more sophisticated than it's going to be," says spokesperson Charles Miller. "This is still in development but it's nothing more than something to make people more aware of what's going on around them, and most people do that now anyway."
Likewise, both the Federal Bureau of Investigations and the Central Intelligence Agency denied roles in any sweeping algorithm to measure citizens' potential terrorist leanings. If anything, the FBI has recently been taken to task for being a tin-cans-and-string Luddite organization. But the FBI is a client of the consumer data aggregator ChoicePoint. And a U.S. official tells the Voice, "Can I categorically deny anybody in government is doing it? No."
An admission that the government is combing through purchase records certainly would help explain why, according to the Naples Daily News, federal agents reviewed the shopper-card transactions of hijacker Mohammed Atta's crew to create a profile of ethnic tastes and terrorist supermarket-shopping preferences.
Algorithms are already used to search for things as diverse as credit card fraud and ideal college applicants. Since 1998, airline ticket buyers have been sifted at the reservations desk by the Computer Assisted Passenger Prescreening System, or CAPPS, a net championed by Al Gore and set to expand dramatically. The group overseeing the algorithm, the Transportation Security Administration, won't comment on what new data might be added to create CAPPS 2.
"At a conceptual level, the work that these algorithms do is not much different than the work that a detective undertakes in assessing whether an individual is a suspect in a crime," explains Christy Joiner-Congleton, CEO of Stone Analytics, a leading developer of such programs. "Good algorithms sort through mountains of outcomes and possible contributing factors and identify relationships for very rare events, like terrorism. The more exotic the outcome, the more data is needed to discover it, and the more sophisticated the algorithm must be to discover it."
Academic mathematicians and statisticians who design algorithms have also called for broader databases. Among them are Kafadar and Max D. Morris of Iowa State University, co-authors of a new paper titled "Data-Based Detection of Potential Terrorist Attacks on Airplanes." They note that "[a]fter the fact, some common elements of the suspected terrorists are obvious: None were U.S. citizens, all had lived in the U.S. for some period of time, all had connections to a particular foreign country, all had purchased one-way tickets at the gate with cash. The statistical odds that five out of 80 revenue passengers (in the case of one of the four hijacked flights on September 11) fit this profile might, by itself, be unusual enough to warrant concern."
Racial profiling finds quasi-acceptance in the hunt for terrorists, as it does in the drug war or the pursuit of serial killers, who tend to be middle-aged white men. But Kafadar and Morris argue that the "historical data must be relevant to a specific flight. For example, a United flight leaving San Francisco for Seoul, Korea, could be expected to carry a much larger fraction of Asian passengers than one might see on a flight from, say, Des Moines to Denver," the authors write. A trip like Atta's, Kafadar tells the Voice, "wasn't a flight coming from Saudi Arabia. There were a disproportionately high number of Arabic names given about 80 people to choose from."
But the algorithm method didn't fail on 9-11the human response did. When the screening program spotted something unusual about at least one of the flights, the people in charge elected only to reinspect the luggage. According to The Wall Street Journal, CAPPS tagged hijackers Nawaf Alhazmi and Khalid Al-Midhar because they'd reserved their tickets by credit card, but paid in cash. The right-wing National Reviewslammed CAPPS for failing to include race, religion, and national origin in its calculations or to tie the system into manual searches of passengers, and not just baggage.