Are Fingerprints Unique?

MattJ writes "There's an incredible article in LinguaFranca about fingerprints. Maybe they're not all unique after all. And maybe those fingerprint experts are not much more scientific than handwriting experts. Fascinating details."

Comments

Comment removed

[account_deleted]

Comment removed based on user account deletion

Well, it's time to do the math....

[ka9dgx]

It's time to push for someone to do the math, some experiments, and determine the actual error rates, factors, etc. I can't believe there is NO possiblity of an error in a "match". There has to be a false-positive as well as a much higher false-negative rate, we need to know what they are, so we, as a society, do the right thing, instead of the emotionally appealing expedient.

To guestimate, I suspect the real numbers are about 10^12:1 false matches, which is quite alarming, if there are 10^7 entries, one in 10,000 will be a false hit with someone totally at random. The false negative rate is probably closer to 1% and might be as high as 10%.

Does anyone have better numbers, after all, I'm just guessing at this point.

--Mike--

Well, I guess...

[Tebriel]

Puddin' Head Wilson was wrong. Bad Mark Twain!

Finger Prints

[CakerX]

Well fingerprints although not unique as we once though still provide good evidence. I remeber reading somewere that about 1 in 10,000 people have similar fingerprints. they still would be crucial evidence if the over 100 people with the same fingerprints in the country never set foot in the city.

$10 Bet...

[the real jeezus]

ever since the mid-1980s, when Automated Fingerprint Identification Systems (AFIS) began to emerge...

Great. We're trusting a multi-variable analytical computer system that's as old as Pac Man... Especially considering that:

The strongly held belief among FPEs that latent fingerprints can be matched to one person alone, wrote David Stoney in a 1997 legal practice manual, is "the product of probabilistic intuitions widely shared among fingerprint examiners, not of scientific research."

and

...proficiency tests reveal high rates of error by FPEs.

I could just picture the conversation in the lab at Lockheed International Conglomerate:
"Richard, I've been thinking...what about distortion from pressure on either the latent print or the test print, or both!" "Oh Sam, you know we only have 640KB of RAM to do our matrix multiplications in...and besides, the government's paying us a lot for this. We're gonna have one hell of a Christmas bonus..."

Consider this, too:

But forensic fingerprint identification is supposed to compare two different impressions from the same finger. As Stoney noted, "No two things, no two representations of this person's signature, no two representations of their fingerprint will be exactly alike." The test, therefore, ought to have used two different impressions from the same finger to establish a baseline score for a match. "It is really shocking," Stoney testified, "to see something presented...that doesn't have that basic element of forensic examination in part of the study."

Great, they may as well be dowsing for water wells now... Okay, I bet $10 that the feebs' computer (AFIS) is a total fraud. The burden of proof, however, is on you...

This sucks. I'm going back to bed.



In 1999, marijuana killed 0 Americans...

Well, why not test it?

[Jonathan]

The software is available (in every police station), the data is available (also in the police stations) -- let's just see what the percentage of cases where the programs claim that the figureprints of two different people are the same.

Just had fingerprinting done

[truefluke]

I needed to be fingerprinted before they would give me my green card, what fascinated me was this machine they took my fingerprints on.

At first glance it looks like a photocopier gone haywire, with a largish monitor. This is where your fingerprint image gets projected. There are several squares of different sizes, and they look similar to the glass partition of an everyday photocopier.

No ink involved, He just asked me to lightly place my fingers on the glass, and he held my fingers at some points to keep them from jiggling (he was very exacting about it).

When he was satisfied, he brought the images of my prints up on the screen in front of us. He remarked "Well, no life of crime for you". He was referring to the fact that I had "excellent patterning". He remarked that "You have excellently defined fingertips" etc. etc. He then showed me the neatest thing he magnified my prints to the point that he could COUNT the skin pores between the whorls of my thumbprint

Actually I was impressed, since up til that point it had always involved me getting my fingers full of that stupid ink and having to roll them around on a piece of paper. That machine was much cooler.

And for the record, I don't have one :)

A double standard?

[ahertz]

Folks, rememeber the OJ trial? There we saw DNA evidience, based on modern science, with proven error rates and a far lower possibility of subjective reading, being totally rejected as forensic evidence. And yet, many people are locked up every day on the basis of fingerprint evidence, based on 19th century science, with unproven error rates and based entirely on subjective judgement.

Does anyone else see a double standard here? And how can we help society to make better, objective scientific judgements, rather than just listening to demagogues?

The birthday paradox

[XNormal]

The first person to fashion a statistical foundation for this assumption was the British gentleman scientist Sir Francis Galton. He calculated the probability that any two fingerprints would resemble each other in all particulars as one in sixty-four billion.

Even if you accept this figure of one in 64 billion the birthday paradox predicts that you can expect to find an identical pair of fingerprints in a database of sqrt(64e9) fingerprints which is just a little over 250,000. I believe the FBI fingerprint database is significantly larger than that. After this "threshold" of sqrt(N) is crossed the number of duplicates starts to rise quite sharply as the database size increases.


----

Re:Fingerprint Seeds

[SEWilco]

Well, it's that process of fingerprint growth that needs to be understood. Without knowing what causes the variations and the probabilities of the various variations, we don't know the odds of two fingerprints being the same.

That's the point of the article -- we don't really know those odds, as all the popular numbers seem to have been created mostly by guessing.

Signal to noise ratio

[RobertGraham]

Of course not all fingerprints are identical. This fact is not in question. The problem is that not all fingerprint MEASUREMENTS are identical, either. This is a standard "signal-to-noise ratio" problem.

This is easily verified with standard scientific practices. Grab 10,000 people and give them a metal ball and have them handle it for awhile (without telling them why), then grab clear fingerprint impressions. Give the balls to examiners whose job it is to grab as many fingerprint impressions as they can. Put all this into a database, then start pulling out fingerprints and having examiners match them up.

What is the error rate whereby fingerprints were matched incorrectly?

As I see it, the real problem is that people can only think in black or white. The focus of this little question has been if fingerprints are unique or not. The average person isn't mentally equiped to think in terms of "How unique are they?", a vast grey area. Science can never answer black-vs-white questions, but they can certainly measure grey.

Read the article, people

[Anne Marie]

The problem isn't whether fingerprints are unique (although it's an interesting point that their uniqueness has never been proved). For all the author cares, they are unique. It's irrelevent to his chief argument.

The problem is that exact complete fingerprints aren't used in forensic investigation. Mere fragments are. And not just fragments; sloppy fragments, read by making impressions with the detective's dust and transfered onto paper. The question is therefore whether those fragments are unique relative to other fragments, given the additional fudge factor in how they're read, and whether the use of fragments instead of complete fingerprints is sufficient.

People need to work on their critical-reading skills.