- Pupil dilation has an impact;
- Contact lenses have an impact;
- Sensor changes (ie, someone has been enrolled on one system and is being matched on another) have a significant impact (even when using the same software);
- Irises change over time more than had been anticipated. The effect on false reject rates is small, but measurable,
In all of the cases, it is the match distribution that is changing: in other words, it’s “fail safe” in that the system behaviour is such that false rejects go up but false accepts do not. So not too bad. But at population scale, the number of false rejects will still be enough be noticeable and dealing with the false rejects effectively (which might mean different things in different environments) will be central to the success of schemes.
Kevin made a terrific point in passing, very clearly explaining to difference between “truth” and “marketing”. He wasn’t alleging any malfeasance, but noting that marketing people don’t understand technology properly, in the particular case of biometrics the disconnect between market and reality is quite high. And can be very misleading, such as when marketing people do not understand what the different kinds of error rates mean, for example. I would go a little further: not only do the marketing people not know, nor do the people who are procuring the system.
As an aside, his comments about the match and non-match distributions in relation to pupil dilation might have explained my negative experiences with the British government’s IRIS scheme at Heathrow and Gatwick. Could it have been that my pupils were dilated by excitement and enthusiasm when I enrolled? Who knows. But i do know that the system — which I no longer use — is a living testimony to the wisdom of Kevin’s point about false rejects. If I decide to try my luck in the IRIS lottery, and am falsely rejected (once this had happened three time I gave up), then I have to go to the back of the queue to for passport control. That doesn’t seem right: there should be a “fast lane” for people who are enrolled in the system (what’s the point of an RFID passport if it doesn’t help here: you could wave your passport on the way into the fast lane and the system could easily check if that passport is enrolled in the system) but rejected.
These opinions are my own (I think) and are presented solely in my capacity as an interested member of the general public [posted with ecto]