Hacker News Reader

Comment by steveharman | Hacker News Reader

> Police in Britain, meanwhile, said a fingerprint check came back positive for the name Bouon Emmanuel Febile, a citizen of Cameroon. Any elation at a breakthrough was muted by a follow up cable from London: It was also positive for a different citizen of Cameroon, a citizen of Haiti, and a man of unknown citizenship.

Err, does this cast doubt on the reliability of fingerprint evidence?

The term "unique" is widely used in conjunction with fingerprints.

View on HN

> Err, does this cast doubt on the reliability of fingerprint evidence?

This isn't really news, though not well publicised in mainstream media.

Fingerprinting isn't really "unique" enough to identify an individual from a very large number of suspects. It is good enough to help identify one suspect from a small pool. (And as soon as you run it against a large criminal database, you've opened yourself up to risking misidentification).

There have been several wrongful convictions based on fingerprint data [0] that were later overruled.

Fingerprinting has been accepted as scientifically accurate, but that hasn't really been verified by studies in the real world. [1][2]

[0] https://www.bu.edu/sjmag/scimag2005/opinion/fingerprints.htm

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3093498/

[2] https://www.hsdl.org/?abstract&did=458960

I bet somebody will make an app at some point to collect everybody's fingerprint, under the guise of finding your "fingerprint buddy" or something.

Under the guise of unlocking your phone...

Fingerprints are a very feeble form of biometrics. India's, Aadhar - A biometric-based ID system that primarily uses fingerprints is the best case study to understand this. Fingerprints, like every other human organ, are subject to wear and tear, this combined with probabilistic algorithms to find a match for a given set of fingerprints makes it a fragile system for "unique" identification.

It's reasonable for casual biometrics, such as a phone unlock, where the likelihood of someone with a match trying to use your phone is remote. Many people will choose the same password or PIN, for example, but of course you can't change your fingerprint on a whim.

In the presence of other evidence, it's definitely useful to narrow down to reasonable doubt territory. Even DNA isn't unique, in the case of twins, but is still used all the time.

Not only twins show the same DNA evidence. Cheap methods for DNA fingerprinting as used in criminal investigations have nonneglibile error bars.

Not just cheap methods - even if the methods are better, they have a history of not being followed correctly [0][1][2].

[0] https://nypost.com/2017/02/10/thousands-of-convicts-told-tha...

[1] https://www.nytimes.com/2003/03/16/weekinreview/the-nation-y...

[2] https://www.kuer.org/post/private-utah-dna-lab-faces-scrutin...

The fingerprints could be matched, but when he was picked up be gave false IDs that were recorded against those prints.

Some quirk meant they weren't initially matched (they took prints, but didn't run a match at that point in time).

Exactly. The article doesn't clarify whether the prints matched multiple people, or the man had given false names and info to multiple countries/systems which might never have been cross-referenced before

One could give false info to Border Patrol and county or state police here in the US and not be caught in certain circumstances. Especially if we're talking about encounters going back a few years or more.

Needs more info

I did some basic research on fingerprints 15 years ago.

Fingerprints have a "birthday problem." The metrics cluster around certain values because fingerprints ARE NOT RANDOM. This is why your gym likely takes your fingerprint and your phone number for access.

Same with DNA actually. There are many, many humans whose DNA appears to be the same to our primitive tests.

You will run into this problem anytime you assume the underlying data is distributed well, but it is not, actually.

If you want to really see behind the matrix, look for "secondary classification". Scary stuff what evidence people can be convicted based upon.

Reply on news.ycombinator.com