Danyel Fisher has a good post on fingerprints and false positives. He doubts the guilt of the Oregon man being held in the Madrid bombings case based on fingerprint evidence - only time will tell for sure, but he does make an excellent point that "...if you are "one in a million," there are 293 of you in the USA..." False positives will become more and more problematic as more people are fingerprinted. One thing Danyel doesn't discuss, but which is extremely important, is that the resulting false positives will be believed accurate with a high degree of confidence, making it much harder to convince authorities that they are indeed false positives.

Also worth checking out is the paper linked to at the end of Danyel's piece: On the Individuality of Fingerprints (pdf).

The problem with this line of argument is that Mayfield isn't just a random match of a fingerprint.

He's an Islamist nutcase. What's the chance that an Islamist nutcase has a random fingerprint match to an Islamist bombing? Not directly calculable, but arguably pretty slim.

I'm not familiar with the details of this case, but I feel like commenting in general.

If the only evidence is a fingerprint (or partial) and there is no other evidence, circumstantial or not, then one could have some doubt. I would not convict in such a case without other evidence. If however, there is corroborating evidence like motive, means, associations, etc, etc, then one can see the fingerprint as conclusive physical evidence.

I don't have a problem seeing the situation either way -- weighing evidence doesn't perplex me, and I don't understand why others have problems making reasonable decisions weighing evidence (the OJ Simpson case comes to mind).

Fingerprints are not 100%, even DNA evidence is not 100%, but a basket of indicative factors including one or both should make the decision rather easy.

A related point worth making is that the decision in legal cases are NOT scientific. These are judgements made with imperfect evidence. The Defense can attack the problem in 2 ways, disprove the allegation (unlikely, but possible) or sometimes just weaken the positive case below the legal threashold of conviction (much more likely).

Physical evidence like fingerprints and DNA are good for finding truth because they tend to be much more reliable than other bits of evidence. This can exonorate or convict a suspect.

(as an aside, there are individuals that have 2 distinct sets DNA in their bodies. A sort of living set of conjoined twins in one body, and not nearly as rare as you might imagine. This does affect DNA tests.)

I won't pretend I read the whole thing, but it looked like the "Individuality" article was primarily focused on automated fingerprint matching, especially related to biometric security scanners. False positives would be a serious concern there.

But, my understanding is that a law enforcement fingerprint search starts with an automatic scan where the expert notes certain key points on the print and the system scans for those. There can be multiple matches, but the final determination is always manual, with a very detailed comparison.

Clearly, it isn't certain that fingerprints are unique, but I've never heard of any examples of people with indentical fingerprints. And I don't know how good their samples are here. But if they are good, and with his background, it looks extremely likely he is involved. With a little more evidence, it should be able to meet the "Resonable Doubt" test.

Whether it's fingerprints or one perfectly formed fingerprint is unclear. One good one and some partials might cinch it.

There's a lot of bad data about fingerprints and fingerprinting and movies and TV Shows like CSI don't help one bit.

There's a lot of potential problems with Fingerprints as a bio metric and always have been, which is why police forces still have to train officers to take good prints and forensic departments still have trained fingerprint officers. Then there are lots of proceduural ways that these differ - many US law enforcement agencies use a "point" system for identification, allowing X number of points of simularity to make a confirmed match. The UK use a "reasonable doubt" system. There is an early court case in the UK where fingerprints were established as a legal identification mechanism, but I don't know the legal status in the US.

With computer systems there are so many areas where variables can be introduced that it does become a concern for systems like the immigration service one.

Top issues there can be include:

1) The sample print quality is important in a many-to-many matching service - too much pressure, for example, will deform the print slightly and lead to errors. It's one of the reasons why officers should hold your finger when taking an ink print.

2) Fingerprints change over time. Anybody who works in a manual job using their hands in a poor environment and so forth will have poor fingerprints for a computer to read. Builders are a good example of this.

3) Many to many matching is a much harder IT problem than single to single - I'm concerned that this system is being brought in before biometric passports are common. This will lead to a number of potential false readings.

4) Partial prints are often abysmal and other data points are used for getting to the right person.

Historically a typical match of a partial print would take around 2 weeks in a typical UK investigation, more manpower could speed that up. I'm not sure what that is now as my data is about 10 ye;ars out of date, but human's still have to work through the final list the computer generates and that all takes time.

poker Have a nice day! :)

poker Have a nice day! :)

http://aqdrfbqfreqes.host.com
desk3
[url=http://aqdsfbqfreqes.host.com]desk4[/url]
[link=http://aqdafbqfreqes.host.com]desk6[/link]