Are Fingerprints Unique? Not Really, AI-Based Study Finds (cnn.com) 60
An anonymous reader shares a report: "Do you think that every fingerprint is actually unique? "It's a question that a professor asked Gabe Guo during a casual chat while he was stuck at home during the Covid-19 lockdowns, waiting to start his freshman year at Columbia University. "Little did I know that conversation would set the stage for the focus of my life for the next three years," Guo said. Guo, now an undergraduate senior in Columbia's department of computer science, led a team that did a study on the subject, with the professor, Wenyao Xu of the University of Buffalo, as one of his coauthors. Published this week in the journal Science Advances, the paper seemingly upends a long-accepted truth about fingerprints: They are not, Guo and his colleagues argue, all unique.
In fact, journals rejected the work multiple times before the team appealed and eventually got it accepted at Science Advances. "There was a lot of pushback from the forensics community initially," recalled Guo, who had no background in forensics before the study. "For the first iteration or two of our paper, they said it's a well-known fact that no two fingerprints are alike. I guess that really helped to improve our study, because we just kept putting more data into it, (increasing accuracy) until eventually the evidence was incontrovertible," he said.
In fact, journals rejected the work multiple times before the team appealed and eventually got it accepted at Science Advances. "There was a lot of pushback from the forensics community initially," recalled Guo, who had no background in forensics before the study. "For the first iteration or two of our paper, they said it's a well-known fact that no two fingerprints are alike. I guess that really helped to improve our study, because we just kept putting more data into it, (increasing accuracy) until eventually the evidence was incontrovertible," he said.
will this study hold up in an court room? (Score:4, Insightful)
will this study hold up in an court room?
Re: (Score:2)
Re:will this study hold up in an court room? (Score:5, Interesting)
Guilt or innocence is not purely based upon fingerprints. Fingerprints are just evidence and needs to be combined with other evidence. If all you need is a fingerprint then many of these court cases would be over and done in a couple of hours. So if the fingerprint is 1 in a million probability of being from someone else, the person is still the key suspect.
This is America too. We have people getting life sentences even when the fingerprints don't match and instead match a known criminal who was in the area. (prosecutors who withold esculpatory evidence should be made to spend some time behind bars)
Re:will this study hold up in an court room? (Score:5, Interesting)
This is America too. We have people getting life sentences even when the fingerprints don't match and instead match a known criminal who was in the area. (prosecutors who withhold exculpatory evidence should be made to spend some time behind bars)
This a thousand times. Blackstone's ratio [wikipedia.org] is a founding concept. Hell, I'd argue death penalty for any prosecutors that willfully withhold exculpatory evidence. That's one of few places a death penalty would actually make a positive difference.
Re: (Score:1)
This is America too. We have people getting life sentences even when the fingerprints don't match and instead match a known criminal who was in the area. (prosecutors who withhold exculpatory evidence should be made to spend some time behind bars)
This a thousand times. Blackstone's ratio [wikipedia.org] is a founding concept. Hell, I'd argue death penalty for any prosecutors that willfully withhold exculpatory evidence. That's one of few places a death penalty would actually make a positive difference.
I agree.
Re: (Score:1)
As any AI scientist knows, there are two dimensions of error in a classification problem, sometimes called type I and type II or sometimes described as false positives and false negatives, or confusion matrices, etc.
https://en.wikipedia.org/wiki/... [wikipedia.org]
A criminal trial is a binary classifier, so there are two relevant forms of error. You cannot specify strong constraints on just one, it can cause the other error to b
Re: (Score:3, Insightful)
sometimes it's better to put a few more innocents in jail, because overall this can help the criminal justice system do its job substantially more efficiently overall. Everything is a balance.
That is double nonsense.
A) the real culprit(s) is still out there
B) you are lying to yourself, insisting you got the right one(s)
Re: (Score:3)
apples vs. oranges (Score:2)
the post was about a system. So it makes no sense to argue based on one single case.
Another example: we're better off if firefighters break the traffic rules - even if it costs a few lives here or there...
Re: (Score:1)
Do not get your point.
So it is oki to put innocents behind bars?
Fooling yourself that the laughing real culprit is out there to do more crimes?
What would be the benefit of that?
Re: (Score:2)
Innocents in jail, culprits roaming free are failures of a system.
If the failures are *extremely* rare, then that system *might* be good.
Re: (Score:1)
In the US the failures are not rare.
Judges and police officers get promoted by convictions.
The most stupid thing one can imagine.
attorney: aiiee!!! (Score:3)
>ometimes it's better to put a few more innocents in jail, because
>overall this can help the criminal justice system do its job
>substantially more efficiently overall. Everything is a balance.
yikes.
This is absolutely, fundamentally and categorically at odds with the US legal system and its principles! (British, too)
While the substance can be found in cases going back hundreds of years further, we know it best in the US as "[W]e are to look upon it as more beneficial, that many guilty persons shoul
Re: (Score:1)
Re: (Score:2)
there's no emotional issue here, and never was.
It is a *value* judgement that protecting the innocent is more important than catching all criminals.
And it's not a two hundred year old freeze; this comes from the Common Law,, which is nearly 800 years old, and relies on even earlier tribal law.
Freedom is important, and isn't going to be given up over a call to mathematics.
And we're better off because of that.
Re: (Score:2)
And if that's the case, the prosecutor has to make sure that the jury understands that the standard of evidence is "beyond a reasonable doubt," not "beyond the shadow of a doubt" as the defense counsel will try to persuade them is appropriate.
Re: (Score:2)
Which is how they've been doing this for many decades.
Re: (Score:1)
Guilt or innocence is not purely based upon fingerprints. Fingerprints are just evidence and needs to be combined with other evidence. If all you need is a fingerprint then many of these court cases would be over and done in a couple of hours.
You're forgetting plea bargains.
Tell the suspect you have fingerprints tying them to the crime (works for DNA too) and trick them into pleading for a lesser charge just so they don't get the book thrown at them. Even if innocent it's often not worth the risk and expense of going to trial.
Didn't even get to the couple hour (5 minute) courtroom.
Re: (Score:3)
Re: (Score:2)
Re:will this study hold up in an court room? (Score:5, Informative)
Re: (Score:3, Insightful)
The 2004 Madrid bombing case already demonstrated to the FBI that fingerprint identification can be suspect.
Flawed fingerprint identification isn't the same as identical fingerprints, what are you smoking?
The case you linked to is interesting, here's a story on it. https://www.nbcnews.com/id/wbn... [nbcnews.com]
"Last week, Spanish authorities said the fingerprints of an Algerian man were on the bag. Jordan said FBI examiners flew to Spain, viewed the original print pattern of the fingerprint on paper, and agreed that it was not Mayfield's."
Why did the FBI have to fly all the way to Spain just to prove the prints didn't belong
Re: will this study hold up in an court room? (Score:2)
Doesn't really matter (Score:5, Informative)
Re: (Score:1)
Nobody has ever been convicted on fingerprint evidence in isolation. It's merely a piece of the picture that also includes other evidence that adds up to a correlation that is highly suggestive that someone did (or did not) do a thing.
Much the same as DNA matching is done on probabilities, but nobody is convicted purely because their DNA was found somewhere, as it's very easy to come up with a plausible reason that injects "reasonable doubt."
Combine DNA, fingerprint, geolocation from phones, trace evidence
Re: (Score:3)
Re:will this study hold up in an court room? (Score:4, Informative)
will this study hold up in an court room?
Not sure it matters. What they're claiming seems to be that fingerprints from the same person are similar and that this may help in cases where the fingerprints found at a crime are from different fingers than on record for a person -- which is rare as all 10 are usually taken and often a palm-print too. In any case, it would be easy enough to simply get *all* the fingerprints from a suspect, assuming they're available, a compare them to those found. I imagine that in cases where current fingerprints aren't available, like if a person is dead (for a while or cremated) , or fingers have been lost (due to an accident, not misplaced :-) ) the validity of the study itself may be of more importance, especially in a cold case.
Re: (Score:1)
Reminds me when Odin once was asked, where he lost his eye, and he answered: I did not lose my eye. I know exactly where it is.
Re: (Score:2)
Re: (Score:2)
The title of the article is misleading. If you click the link to the actual article, the subtitle says this:
Columbia engineers have built a new AI that shatters a long-held belief in forensics–that fingerprints from different fingers of the same person are unique.
This study did not prove that one person's fingerprints might be shared by other people.
Well known (Score:2)
This has been well known for decades at this point - fingerprints aren't unique, but they've been used by prosecutors since inception to close cases. This is going to throw those gates wide open and "burden" the courts with a bunch of appeals.
The science on fingerprint uniqueness was never good, but it won cases, so they kept mum about it. Its advocates knew this, unfortunately, much like the vast majority of "forensic police work" techniques. It's why I made a quick course correction in school and avoided
Re: (Score:2)
Very few things are "unique". Not in physics, and especially not in the courtroom. Even when this idea that fingerprints were unique was floating around, many people didn't believe it. However fingerprints do provide very good evidence to find a suspect, and good enough to hold up in court when combined with other evidence.
Even DNA is not unique in the courtroom, they're not investigating each and every nucleotide to make a match - because those don't even match up within the same human. Instead they're p
Re: (Score:2)
Slightly disagree.
The level of proof needed for a partial fingerprint "match" is tenuous, at best. Most fingerprints are only partial prints, and they're often smudged, stretched, and difficult to "lift" (ie get a good copy of the print onto paper/tape). Even then, what counts as a "match" is usually just 2-3 features on the fingerprint in common (because that's all that's typically visible).
So when they say "the fingerprint matches", it's usually the rough equivalent of saying 2 people are the same height,
Re: (Score:3)
We don't need AI to tell us this is a problem it was known for years. It's a statistics game.
The FBI was quietly upgrading their matching techniques because the odds turned out to be way higher than they expected... I don't remember but it was something like 1 in 2 million or something on that level. Not good enough when you have a crime in a city where millions of people pass thru the city; if not have millions living in the city itself. Back 100s of years ago (early fingerprinting) there was only like 400
Re: (Score:2)
However... Probabilities stack nicely. If there is evidence connecting you to a crime AND they find prints that match yours, the odds are extremely close to 100% they're yours.
This is pretty much the way it's supposed to work - at least before everyone started watching CSI and adopted the fantasy that forensics can absolutely prove much of anything. You look at all the evidence and it is the collective probability you use to judge fact from fiction. And you're never truly 100% sure but you can get close
Re: (Score:2)
>"This is pretty much the way it's supposed to work"
Exactly. There are likely no convictions based on NOTHING but circumstantial evidence, which is what a fingerprint is. They need to show motive, opportunity, and usually some type of witness (person, video, audio, etc). And that might be combined with other circumstantial evidence (DNA, blood type, dental records, hair matching, financial records, etc).
Whole fingerprints, in ultra-resolution, probably are as close to "unique" as anything can be. In
Re: (Score:3)
Two things:
1) Lifting good prints from a crime scene is nowhere near as common as you are led to believe by movies and television. If anything is found, and in a condition to be lifted, it's usually both partial and somewhat smudged.
2) While it is true that it isn't exactly rocket science to plant prints, what you (and most cops and even forensic labs) don't know is that the oil in fingerprints is distinctly human and even contains identifiable DNA fragments. Nobody is getting a copy of your prints (alre
Re: (Score:1)
but they've been used by prosecutors since inception to close cases. This is going to throw those gates wide open and "burden" the courts with a bunch of appeals.
No it won't. A prosecutor doesn't care if the index finger or the middle finger is confused, and all this study is doing is affirming that the same person can have the same fingerprint on two different fingers.
Also unique isn't necessary in a court room. The burden is beyond reasonable doubt. The odds of being incorrectly convicted due to fingerprints when you don't have a twin and you are both so in cahoots with each other are infinitesimal. People aren't convicted on fingerprints being unique on the plane
Article Title is MUCHO importante here: (Score:5, Informative)
Columbia engineers have built a new AI that shatters a long-held belief in forensics–that fingerprints from different fingers of the same person are unique...
Re: Article Title is MUCHO importante here: (Score:5, Informative)
Re:Article Title is MUCHO importante here: (Score:4, Insightful)
That's actually far more interesting -- they found that the prints of a person's various fingers are not independent variables. This should be useful for when prints are partial or smudged or otherwise low quality.
This is clickbait (Score:3, Informative)
I would bet... (Score:1)
1) They're unique between individuals (or probably unique to such a high degree the difference is semantics), so long as you are using an adequate resolution for your comparison method. Police and digital fingerprinting systems may (or may not) do this to a standard we should feel comfortable with.
2) Fingerprints from different fingers of the same person should be highly similar because the human body grows fingers from the same plan at the same point with minor position and length adjustments. And you ca
Conviction by assertion (Score:2)
Fingerprints have not been shown (cannot be shown?) to be unique. Compounding this problem is that, at least sometimes (I remember a UK case where the police were trying to frame a fellow policewoman), fingerprints are examined for similarities only. Differences are ignored.
Not constant (Score:3)
Re: (Score:3)
Why AI? (Score:1)
To prove that fingerprints are not unique, we don't need "AI-based studies". We need to find two different fingers with identical fingerprints.No more, no less.
Fingerprint readers (Score:2)
Finally I can dump my MacBook Pro fingerprint reader that only works half the time anyway.
define "unique" (Score:2)
It's been known for a long time that if "unique" is defined by the n special locus points used for fingerprint matching, that fingerprints are far from unique. There was in fact a famous case of a guy in WA who got caught by fingerprints yet was provably nowhere near the crime, like 10 years ago or so.
If you define "unique" relative to a real number then "unique" depends on how much variation you will accept.
So, like many things, if you have enough data, you can estimate the probability that a fingerprint
Doesn't this invalidate many legal cases? (Score:2)
Given that the American judicial system operates on the fundamental basis that. people are innocent until proven guilty by definitive fact and reasonable lack of doubt, and there has been a reasonable doubt cast on all fingerprint matching, Wouldn't that mean that fingerprint matching is no longer legal as a basis for guilt in America now that we've found the scientific basis for that?
And wouldn't that mean that people who were put away based on evidence around fingerprints might have actually been innocent
Re: (Score:2)
Not likely. The subtitle of the linked article says this:
Columbia engineers have built a new AI that shatters a long-held belief in forensics–that fingerprints from different fingers of the same person are unique.
This study did not conclude that one person's fingerprints might be shared by another person.
According to my HP laptop (Score:2)
My fingerprint on my right index is the same that my son on his ring finger