Airport security officer stops to let dog sniff luggage (pic: courtesy of USA Today)

Airport security officer stops to let dog sniff luggage (pic: courtesy of USA Today)

11/20/2007

Practice Random Acts of Security

by Lauren Cox

From the Sesame Street “one of these things is not like the other” game to SAT math questions, we are trained to find patterns. It's a difficult task and it's why code crackers at the National Security Agency get paid so well. But recognizing randomness—phenomena that have no pattern—can be just as difficult, and essential to police and airport patrols.

People create and recognize patterns by nature and there is a good chance that your family and coworkers can predict them: how you stir coffee, how you start your emails, how you greet clients over the phone. But when a police officer falls into a routine at his patrol post, it's a criminal, not a coworker, who might predict his behavior—and find the perfect time to attack.

Although airport security officials try to patrol randomly to deter planned attacks, as humans they are prone to unconscious patterns. To solve this problem, officials at Los Angeles International Airport (LAX) have started testing a computer program that uses game theory to generate random patrols daily.

Ph.D. student Praveen Parchuri wrote the randomized  algorithm computer  program for his thesis at the University of Southern California. The trick is to make the patrol as random as possible while paying attention to risks unique to the airport.

Imagine three houses on a block. House X and Y are nice, but a tax man lives in house Z and has the social security numbers for thousands of taxpayers on his desktop at home. No one wants to be robbed, but house Z should be patrolled twice as much because it has more value. If police try to randomly patrol this neighborhood, with house Z getting twice the surveillance of house X or  Y, they will most likely form an unconscious pattern. By using Parchuri's game theory program, officers can generate an effective patrol sequence that will keep house Z under more watch remain random to keep a robber from predicting when the cops will show up.

Computer programs can beat chess champions, but it's not certain whether game theory  algorithms can beat the criminal mind. Sergeant Jack St. Hilaire works for Boston University and is about to finish a Ph.D. focusing on patrolling methods. He cautions that police have been testing patrolling methods since the '70s and still haven't found a magic bullet. He said  that using random patrolling alone can't catch criminals.

More than eight major police theories have come and gone since police started questioning their patrolling methods. In 1974, the Kansas City Police Department did an experiment to see if routine patrols prevented crime. They found that if they increased the number of  marked cars, decreased the number of marked cars or patrolled in unmarked cars the crime rates still remained the same.  A 1982 article in The Atlantic Monthly brought the “broken window” patrolling theory into the public sphere. According to the theory, people commit crimes in neglected and poor neighborhoods because they assume the residents already put up with criminal activity. The theory contends that if communities clean up the neighborhood and police patrol neighborhoods for littering and minor offenses, then the crime rates will drop. Rudy Giuliani adopted a form of this theory when he was mayor of New York City and to this day the media debates whether it worked, or whether the crime drop was a result of a coincidental decrease in the crack-cocaine epidemic.

A patrolling theory that has been proven successful, according to Sgt. St. Hilaire, is the “cops on dots,” or COMPSTAT, approach: crunch numbers of crime statistics on when, how and where crimes are committed and send police to stem the trend. Parchuri's algorithm resembles COMPSTAT  because it assigns “cops” to the right patrolling “dots,”or weak points in the airport.

So step aside, RoboCop. The new science and technology of policing isn't necessarily super-human weapons but the scientific method: gathering statistics, forming a hypothesis and testing it. This year, LAX is putting randomized patrolling to the test.