First of it, fprint works reasonable on my hp 2510p Notebook, Thanks.
But still, I get a lot of rejected validation. I suspect that taking
only one one fingerprint for enrollment could be the cause. There is no
good way to rate the reproducibility and quality of one single scan. So
I tried to validate this thesis by just enrolling the same finger
multiple times (well the entire left hand scans are now right index
fingers :)), voila. this fixed it, the recognition is now much better.
But well this is only a bandaid fix. It would be nice if normal
enrollement would take serveral swipes, compare them and store few
sufficent ones, lets say take 5-10 swipes, find a median, throw away the
ones which are really out of bounds (bad swipes), throw away the ones
which are very similar (don't improve accuracy) and then store for
example 3 scans which reflect the range of valid recognitions.
Another idea is that pam_fprint should allow one to make few tries
before giving up. Since I have a password query after a 'sufficent'
fprint scan, a failed attempt nags with a password query instantly.
Having a chance to try again within one fprint authentication try
('retry=3' option) would make it somewhat more convinient.
Finally some complete other thing: How about logging any failed
fingerprint scans? Then one has the fingerprints of anyone who just
tried to figure it out stored on his system. This is good for forensics
and when this is a well known feature, people won't even try to play
with fingerprint sensors where they should be not scaring the curious
ones away.
Christian
_______________________________________________
fprint mailing list
[email protected]
http://lists.reactivated.net/mailman/listinfo/fprint
