Comment Re:I don't worry (Score 5, Insightful) 503
> My fingerprints are in at least one government database
> (for non-criminal reasons). It doesn't bother me.
> When I decide to rob banks or kill people, then I'll be worried.
Remember that fingerprints in the database are stored as encoded strings describing the location of some branching points in the ridge patterns. Essentially a hash function.
Note that the data on which the hash is constructed is subject to scaling and rotation of the captured image. Note that selection of the points is hard in some people's patterns (because there may be unusually many branch points.
Now note that we are addding a very large amount of new prints to be processed, of which a much lower proportion will ever be needed than in the previous population of the database. There will be extreme pressure to do this quickly and cheaply with less-skilled operators. This will lead to many false matches.
We already have many cases of false matches leading to arrest of innocent people when fingerprint data is shared between FBI and Interpol (made worse by some differences in technical standards between different police organizations. And because most of the victims of these false positives will not be US voters, fixing the problems will not be a high priority.
If you really believe that mass processing of huge fingerprint databases is feasible with acceptably low error rates, you should advocate that a full set of prints for the FBI database should be taken with every US driver's license application. This would have enormous benefits if every fingerprint found at a crime scene could easily be matched. By raising the chance of solving crimes by an order of magnitude, it would create an enormous incentive for people not to commit crimes. But I don't know anyone who trusts the system enough to want this to be done.
> (for non-criminal reasons). It doesn't bother me.
> When I decide to rob banks or kill people, then I'll be worried.
Remember that fingerprints in the database are stored as encoded strings describing the location of some branching points in the ridge patterns. Essentially a hash function.
Note that the data on which the hash is constructed is subject to scaling and rotation of the captured image. Note that selection of the points is hard in some people's patterns (because there may be unusually many branch points.
Now note that we are addding a very large amount of new prints to be processed, of which a much lower proportion will ever be needed than in the previous population of the database. There will be extreme pressure to do this quickly and cheaply with less-skilled operators. This will lead to many false matches.
We already have many cases of false matches leading to arrest of innocent people when fingerprint data is shared between FBI and Interpol (made worse by some differences in technical standards between different police organizations. And because most of the victims of these false positives will not be US voters, fixing the problems will not be a high priority.
If you really believe that mass processing of huge fingerprint databases is feasible with acceptably low error rates, you should advocate that a full set of prints for the FBI database should be taken with every US driver's license application. This would have enormous benefits if every fingerprint found at a crime scene could easily be matched. By raising the chance of solving crimes by an order of magnitude, it would create an enormous incentive for people not to commit crimes. But I don't know anyone who trusts the system enough to want this to be done.
