I wrote:

..... But they seem to confirm them: whole-DB-match was done only for
fingerprint *enrolment*, to check that similar prints hadn't already
been enrolled, while verification was limited to 1-to-1! For iris,
Daugman's encapsulation seems to mean that the trial was 'forced' to do
1-to-many - but note that there weren't 'many' even at the end of the
trial, and only the last enrollee was matched against all the others!

It wasn't quite as daft as that, on full reading. They preloaded the
fingerprint DB with 1.1million dabs from wrong 'uns (actually, they'd be
*mainly* wrong 'uns; but since a couple of years ago prints are now kept
for anyone *arrested*, regardless of whether they were charged or
convicted: http://www.policereform.gov.uk/docs/...ng_mar044.html);
and the iris DB with 110,000 iris scans. So for fingerprint and iris,
enrolment checks were done 1:many against a 'large' number of
pre-existing should-definitely-be-non-matching biometrics; though in the
case of iris it was one-thousandth the size of the proposed DB (we'll
have 100 million iris scans from 45 million residents + few million
longer-term visitors), and a similar ratio for dabs (there seem to have
been 10 dabs/person, so 50million people = 500million dabs, 1.1mill is
one-fivehundredth of that size).

Stefek