FINDING THE FACE OF TERROR:
Using Math to Help Track Madmen
Q&A with Andrew Rukhin, Professor
UMBC Department of Mathematics and Statistics
The emerging field of biometrics technology uses optical scanning among other technologies to rapidly identify individuals based on biological traits such as fingerprints, unique patterns in the eye or face recognition.
Earlier this month, New Zealand's government announced that all of its citizens will be issued biometric passports by next summer. Each passport will include a microchip containing a digital photograph, eye coordinates and an electronic version of the information printed on the passport. Many other nations whose citizens previously were permitted visa-free entry to the U.S. will soon be following suit.
UMBC faculty researchers are working with U.S. government agencies to improve high-technology methods such as biometrics to detect terrorists and their tools of destruction. Through an anti-bioterror grant from the National Institute of Standards & Technology (NIST), UMBC Statistics professor Andrew Rukhin is working to improve facial identification software that could help identify terror suspects at border crossings, transportation hub, and other sensitive locations.
In addition to his work at UMBC, Rukhin is a Mathematical Statistician in the Statistical Engineering Division of NIST, a Fellow of the Institute of Mathematical Statistics, a Fellow of the American Statistical Association and an editor with several scholarly journals. He has published more than 160 papers, given more than 140 invited lectures at various universities and presented about 70 invited papers at statistical conferences.
UMBC Research News spoke with Professor Rukhin to briefly discuss the state of biometric technology and how it could help law enforcement locate and capture terror suspects.
How far away is the U.S. from requiring biometric ID for its citizens?
It is very difficult to judge this. It would rely on the state of commercial algorithm systems, which remain proprietary. Also, political considerations that must be addressed introduce a large element of uncertainty.
According to news reports, there are still many flaws in facial recognition software currently on the market or in testing. What makes faces harder to recognize than fingerprints or the iris of the eye?
Faces are complex when considering differences that arise from changes in orientation or changes in expression. They are also easy to camouflage. Keep in mind that when comparing partial fingerprints against national databases, there is at least a 10% error rate, so there are algorithmic challenges with all biometric methods.
In lay terms, how can mathematics help make facial biometrics more accurate?
By developing more efficient methods for representing human faces in multidimensional parameter space, and improving face comparison methodologies.
How did you first become interested in biometrics as a mathematical challenge?
By consulting with NIST scientists and reading some popular literature on the topic. My interest developed before 9/11.
How are NIST and other government agencies working to balance homeland security with citizens' concerns about civil liberties and privacy?
This is a very difficult question for society, but I am sure it will eventually find a balance.