Biometric template protection and biometric privacy (first half) + Brief introduction to other ML research projects (second half)

As the use of biometric technologies increases, so does privacy worries
about the misuse of biometrics (fingerprints etc.) stored in central databases.
In the first half of the talk, I will introduce the biometric privacy problem
and present some proposed template protection mechanisms from the literature.
In particular, I will talk about our suggestion for layering multiple biometrics
of an individual (e.g. two fingerprints or a fingerprint and a voice pass-phrase)
in one multi-biometric template. The method makes it infeasible to separate the
raw biometric data and difficult to link multiple biometric databases, to track an
individual.

In the second half, I will talk about some of my other research projects very briefly
so as to present my background better for potential collaborations during my sabbatical
stay at University of Michigan. This part will include work on plant identification from
photographs and identifying visual words using text and taxonomy, and an optimization
for Error-Correcting Output Codes.
Berrin Yanikoglu received her Ph.D. degree in Computer Science from
Dartmouth College, USA in 1993. After working on handwriting recognition
and document understanding problems, at Xerox Imaging Systems and
IBM Almaden Research Center, she has joined Sabanci University, Turkey
in 2000 where she is currently a full professor. Dr. Yanikoglu's research
interests lies in the areas of pattern recognition and machine learning with
applications to image understanding and text analytics. She has received
first/second place positions in several international signature verification
competitions and plant identification campaigns along with her students and colleagues.