Professor Ross Anderson: I was assuming that we could use, for the sake of argument, the Metropolitan police’s claim that the probability of a misidentification when you match on 16 points is 1 in a billion. I do not believe that it is that good. It is perhaps in the tens of millions, but that is something on which research is ongoing. The problem that you get in translating a figure like this to population against population matches is that if you match 50 million people in England against 100 million people arriving at Heathrow in four or five years’ time, the number of identifications that you are doing, assuming thatyou are scanning one finger of each person, will be50 million x 100 million. So it will be 5 x 1015, if my mental arithmetic serves, which is an awful lot more than the one in a billion error rate that even the Met would have us believe. If the error rate, given the equipment that we are going to use, is actually closer to one in 10 million, the number of false matches that you will have will simply swamp the system. That is why, for population-screening scale biometrics, iris codes are much better than fingerprints.

See this item in context