[10308] in cryptography@c2.net mail archive
Re: Fingerprints (was: Re: biometrics)
daemon@ATHENA.MIT.EDU (lynn.wheeler@firstdata.com)
Mon Jan 28 17:44:03 2002
To: ji@research.att.com
Cc: cryptography@wasabisystems.com
Message-ID: <OF05CF1CF1.8A0E0329-ON87256B4F.0076F1B1@internet.ny.fdms.firstdata.com>
From: lynn.wheeler@firstdata.com
Date: Mon, 28 Jan 2002 14:54:57 -0700
MIME-Version: 1.0
Content-type: text/plain; charset=us-ascii
I believe NIST published something about FBI needing 40 minutia standard
for registration in their database.
On tv you see these things about lifting partial prints and then sending
them off to FBI to try and find who the partial print matches with, aka the
FBI better have rather detailed recording of whatever part of the print
that happened to be lifted.
That is significantly different than trying to repeat scans in the same
way, on nearly identical surface, from the same angle, of a "full" print
etc. and approx. match at least a minimum number of points. By comparison,
the fbi might need to have higher number of point match based on only a
very specific subarea. That would imply that the needed resolution of valid
points on the minimum acceptable sized subarea equivalent to typical
matching of a full fingerprint.
lets say that FBI wants to do acceptable minutia match on a 15 percent
finger subarea (pure conjecture on my part, i've never even read anything
about minimum resolution needed in partial print search) ... then
presumably the (fbi's) total finger resolution (recording) might need to be
six times higher than a straight-foward comparison involving always
matching a full-finger to the same full-finger recording using essentially
the same methodology each time.
Even at that, the straight-forward fingerprint match (as opposed to the
partial print search problem) is frequently subject to greasy & dirty
finger problems.
ji@research.att.com at 1/28/2002 1:46 pm wrote:
Last week I had to go to my local INS office to get fingerprinted
(part of the green card process is getting your fingerprints OK'ed by
the FBI (and also presumably stored for future reference)). The
process is computerised, with a low-res scan of all the fingers taken
once, and then each finger is individually rolled and scanned on a
much higher resolution scanner.
The process took about 20-30 minutes; each finger had to be wiped
with some cleaning fluid, the glass on top of the scanner also had to
be wiped between scans, and a fingerprinting technician had to roll
each of my fingers with the right amount of pressure to get a clear
image of the fingerprint. Even with immediate feedback on a large
screen showing the fingerprint and how good the scan was, some fingers
took as many as five tries to get an acceptable fingerprint.
Now, this was a special-built device whose only purpose is to scan
fingerprints, operated under ideal conditions by a trained
technician. Draw your own conclusions about the effectiveness of
mass-produced fingerprint scanners that would be integrated in other
devices.
/ji
--
/\ ASCII ribbon | John "JI" Ioannidis * Secure Systems Research
Department
\/ campaign | AT&T Labs - Research * Florham Park, NJ 07932 * USA
/\ against | "Intellectuals trying to out-intellectual
/ \ HTML email. | other intellectuals" (Fritz the Cat)
---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to
majordomo@wasabisystems.com
---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo@wasabisystems.com
