Post-9/11, the U.S. government instituted the U.S. Visitor and Immigrant Status Indicator Technology (US-VISIT) Program, which compares two index fingerprints from every foreign visitor entering the U.S. against a watchlist of fingerprints from several million criminals and suspected terrorists. In the first part of this talk, we develop a new probabilistic model for fingerprint matching that allows for population heterogeneity in fingerprint image quality, calibrate this model using data from the National Institute of Standards and Technology (NIST), and embed the model into a Stackelberg game, in which the U.S. government chooses an optimal biometric strategy to maximize the detection probability subject to a constraint on the mean biometric processing time per legal visitor, and then the terrorist chooses his fingerprint image quality to minimize his detection probability. We predict that switching from a two-finger system to a ten-finger system would increase the detection probability in this game from 0.526 to 0.949. This work was the basis of Congressional testimony last fall, and the Department of Homeland Security recently announced that they are switching from a two-finger system to a ten-finger system. In the second part of this talk, we use new data from Cogent (the biometrics vendor for the U.S. Visit Program) to derive a two-stage, two-finger biometric strategy that works as well as a one-stage, ten-finger strategy. The second stage of this two-stage strategy employs texture matching rather than the traditional minutiae matching.
Manas Baveja is a doctoral candidate in the Institute for Computational and Mathematical Engineering at Stanford University and a CISAC science fellow. His doctoral research is focused on quantitative modeling of homeland security projects.
Lawrence Wein is the Paul E. Holden Professor of Management Science at the Graduate School of Business, Stanford University, and an affiliated faculty member at CISAC.