biometrics

[Rick Smith at Secure Computing]

The essential problem I've always seen with biometrics (and one that 
Dorothy Denning acknowledged in her recent op ed piece without seriously 
examining) is the question of whether it's as efficient to deploy and 
manage biometrics safely as it is to deploy and manage some keyed 
alternative like smart cards or other tokens.

Once you start embedding crypto secrets into your biometric reader, you are 
no longer managing biometrics. You're now managing BOTH biometrics AND a 
bunch of crypto keys. Why not just save yourself the administrative 
headache, deploy tokens, and use that crypto key for authentication?

I'm sure there are applications where biometrics make sense (ATMs, door 
security, and other closed systems like that) but I just don't see them 
working in an open system where your main problem is to associate the 
endpoint with a person. If you also need to separately authenticate the 
endpoint, and that's what everyone recommends, then the system costs go up 
even more.

My favorite biometric implementation is the "fingerprint as PIN" token, 
which several vendors make. There's the Sony Puppy, a credit card 
calculator sized token with a USB cord and an embedded public key pair. 
There are also various PCMCIA readers that (apparently) you can plug in to 
your laptop to provide a biometric lock.

My impression, however, is that these readers provide a PIN-like resistance 
to attack. Once you've cranked the false rejections down to the point that 
it's convenient, the false positives are approaching PIN levels (2^13 
guesses on average).

A nice feature of the "fingerprint as PIN" tokens is, of course, that the 
print never leaves the card. You still have to worry about images of 
fingerprints or rubber fingers, of course. The print is a back-up for 
physical possession.


Rick.
smith at securecomputing.com            roseville, minnesota
"Authentication" in bookstores http://www.visi.com/crypto/




---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at wasabisystems.com