Smugglers know a hundred tricks. They hide drugs in propane tanks, tuck them into high-top sneakers, conceal them in children's toys, and stash them in tins of butter. One smuggler even encased his cargo in tombstones. And every day the smugglers invent a brand new ploy. MITRE is creating a simulation that models the smugglers' evolving tactics in order to help U.S. agencies think like the smugglers—and outwit them.

According to a 2002 study by the U.S. Drug Enforcement Agency, Mexico is the transit point for 70 percent of the cocaine sold in the United States, the producer of seven metric tons of heroin per year, and the primary supplier of marijuana sold in this country. To stop this flood of drugs across the southwest border, agencies such as Customs and Border Protection use screening methods to determine which few people among the hundreds of thousands who cross from Mexico into the United States each day are smugglers. Once smugglers are caught, their methods are studied and incorporated into a set of standard screening techniques. But what of the smugglers who are not caught? What methods are they using to evade detection? MITRE is developing screening strategies to thwart these unknown methods.

MITRE has created a simulation program capable of modeling the complex adaptive behaviors of the smugglers. The program will be capable of generating a steady stream of novel smuggling techniques so that strategies can be crafted to combat them. Once an agency produces these strategies, it can feed them into the simulation to test how smugglers may adapt their methods in response. In just a few simulation runs, the user can gain insight into smuggling and prevention techniques that years of trial and error studies in the field may never have revealed.

The Modeling Complex Adaptive Behavior Project is headed by Daniel Venese and sponsored by MITRE's Center for Enterprise Modernization. Venese previously led MITRE's efforts for the Defense Information Systems Agency's Anti-Drug Network program, for which he designed counter-drug strategies, applied data mining to land border targeting, and developed a variety of advanced prototypes. Setting out on this new project, Venese realized that his findings could extend beyond drug smuggling. "This simulation technology has taken on an added importance with the focus on homeland security. Many of the same challenges faced in anti-smuggling apply to anti-terrorism as well."

Like terrorists, smugglers rely on ingenuity to escape detection. "Major smuggling organizations are highly sophisticated: they observe what's going on, they learn from their mistakes, and they improve and evolve over time," says Venese. "If you didn't know it was a smuggling operation, you would think it was a Fortune 500 organization." To model the complex adaptive behavior of the smugglers, Venese and his team programmed the simulator with an understanding of reinforcement learning, multiple strategies, randomization, and emergent behavior, which are described below.

• Reinforcement learning is when subjects learn from their successes and failures. For example, if the subject determines a tactic is successful, he or she will employ that tactic more often. Faced with a failure, the subject will employ that tactic less often.

• Multiple strategies describes the process of employing more than one tactic so as to avoid putting all of your energies into one effort.

• Randomization describes the smugglers' efforts to avoid being predictable. This involves changing tactics frequently.

• Emergent behavior is when complex behaviors arise from simple actions. Simulators study this phenomenon by programming multiple models with simple actions and then observing the complex behaviors that result from the models' interactions. For example, animal behaviorists used this technique to uncover why geese fly in a "V" pattern. They programmed their model "geese" with a simple understanding of aeronautics. When allowed to "fly," the geese quickly adapted the "V" formation as the most economical mode of flight.

Venese's team soon found that as complicated as it was to program the simulator, running the simulations was even more challenging. The team decided that it needed to incorporate the maximum amount of detail into the border crossing model to come up with useful results. "For example," Venese explains, "one million vehicles cross over a medium-sized border crossing in one year. To model such a crossing at a high degree of fidelity, we had to simulate events lasting as few as two seconds. When you program in those kinds of numbers, one run of one year was taking 80 hours, and we wanted to do 30 runs to get a statistical distribution. So to completely explore the parameters of the simulation we were looking at a 45-year project."

Venese went searching for help to speed up the simulations. He found it right under MITRE's roof. "This is an example of the wonderful synergy you have here. We found out that there was another MITRE team developing a high-performance simulation engine called the JAVA Meets Simulation. Using their technology, we were able to take that 80-hour run down to one hour, and we're still making improvements."

With the simulator now providing information at a reasonable rate, the team could begin collecting information that can be used to help capture even the slipperiest smugglers. The simulator quickly illuminated the most crucial border screening factors. "In any environment, you're always trying to identify the factors that have the greatest effect on your performance—positive and negative," Venese says. "That way you can avoid spending a lot of time and money studying factors that don't have much effect either way."

The Modeling Complex Behavior team can now glimpse the world through a smuggler's eyes. The results have been so successful that MITRE has extended funding for the research project for a second year. Venese hopes that his team's modeling work will be used by MITRE sponsors that are trying to puzzle out—and combat—the behaviors of terrorists and other criminals.

—by Christopher Lockheardt

Related Information

Technical Papers and Presentations

• Java Meets Simulation [PDF, 877KB]

Solutions That Make a Difference.®
Copyright © 1997-2013, The MITRE Corporation. All rights reserved.
MITRE is a registered trademark of The MITRE Corporation.
Material on this site may be copied and distributed with permission only.

Privacy Policy | Contact Us