[LINK] SMH Blurb tries to help ailing biometrics industry - 2

[Stephen Wilson]

Roger,

Excellent critique, well done.

I think the notion of "uniqueness" should be one of the prime targets 
for critics.  The word is used with scandalous abandon.  It's a furphy.

You wrote:

> ... "Biometrics involves capturing information about something unique
>     to an individual - their voice, face, ..."

Unique?  Even the most cursory examination of biometrics in practice 
leads people to the common product specs of False Detect Rate and False 
Reject Rate.  Typically these are around 1 or 2 percent.

Why don't more people -- journalists especially -- spot the clanger? How 
can they be "unique" on the one hand, and suffer one-in-a-hundred errors 
on the other?


>     Neither voice nor face are unique to an individual.

Add fingerprints too.  See the work of Cole in the UK:

"Although conventional wisdom since the nineteenth century has accepted 
the doctrine that no two fingerprints are alike, no one has really 
proven the proposition's validity."
The Myth of Fingerprints: A forensic science stands trial, Simon Cole, 
Lingua Franca 10(8) pp 54-62, 2000; 
http://fp.bio.utk.edu/evo-eco/resources-this_semester/Cole-fingerprints.pdf. 



>     [Variability] is true of biometrics generally, although with biometrics
>     that are based on a unique physical feature, the reasons have less to
>     do with variation of the physical characteristics and more to do with
>     the wide range of technical and operational challenges involved in
>     taking measurements.

To pick up your point on measurement difficulty. So true!

When one looks closely at iris scanning (no pun intended) the impact of 
measurement is clear.  The inventor of commercial iris recognition, John 
Daugman, likes to claim that the chances of two peoples' irises matching 
is one in ten to the power of seventy eight.  The denominator is truly a 
fantastic number -- much bigger than the number of atoms in the universe.

But the practical performance of iris measurement systems is affected by 
lighting, grime, sensor resolution, filtering, digitisation and god 
knows what else.  And when you compare scans taken with different 
cameras, the precision goes right off.  See 
http://www.cardtechnology.com/article.html?id=20050811DQB2XGLL.

In practice, false match rates with iris technology can be 0.001%

That's 10 million million million million million million million 
million million million million million times worse than Daugman 
implies.  That's a precise figure, I'm not exaggerating, he is.

Cheers,

Steve Wilson.



Lockstep
www.lockstep.com.au
-------------------
  * Lockstep Technologies: ICT Secrets of Innovation Finalist 2007
  * Lockstep Technologies: Anthill / PwC Cool Company Finalist 2007
-------------------
Lockstep Consulting provides independent specialist advice and analysis
on authentication, PKI and smartcards.  Lockstep Technologies develops
unique new smart ID solutions that safeguard identity and privacy.