I don’t know about you, but I look forward to the day when I
can stop memorizing passwords and use biometric identification (BI) systems
instead. Unfortunately, iris BI may not be the ideal solution. According to
Samuel Fenker and Kevin Bowyer of the University of Notre Dame, iris BI systems suffer from a phenomenon
called ‘template aging’. That is, the inaccuracies in matching
the enrollment (template) scan to subsequent scans mount as time passes.
To be
clear, these are increases in percentages rather than actual numbers. In other
words, if the false reject rate of pictures taken a month apart was one in a
million, then three years later that same person would be rejected by a scanner
two and a half times out of every million. If this sounds like nothing to worry
about, remember that if iris BI systems became widespread with millions of
people using the device, there will be many false rejections every day.
In the case of iris scans, accuracy begins declining
almost immediately and rapidly gets worse. After only one year, the false
reject rate (the rate at which the system fails to accept that an iris
scan is from the same person) went up by 25-60%. After three years, the
false reject rate went up by 150%.
So what causes template aging? The most obvious source of this problem is that a person’s physical attributes have shifted in some way.
Contrary to popular opinion, irises do change as people age. However,
Bowyer points out that other causes, such as deteriorating imaging equipment or
even posture and environment (which might influence how dilated a pupil is) can
also attribute to template aging. It's difficult to reconcile the short time
frames at which template aging is observed with any of these processes though.
Do people's eyes change that rapidly? Do they choose progressively odd positions
in front of the scanner as time passes? It's hard to say what exactly is going
on. In any case, it's important to be aware of the problem.
No comments:
Post a Comment