Hacker News new | comments | show | ask | jobs | submit login
[dupe] Police extracted fingerprints from WhatsApp photos to convict drug dealers (bgr.com)
75 points by artsandsci 64 days ago | hide | past | web | favorite | 29 comments



(FYI, this was posted yesterady -- there were some interesting comments: https://news.ycombinator.com/item?id=16842414)


I guess law enforcement doesn't pay attention to stuff like this:

"A new American Association for the Advancement of Science (AAAS) working group report on the quality of latent fingerprint analysis says that courtroom testimony and reports stating or even implying that fingerprints collected from a crime scene belong to a single person are indefensible and lack scientific foundation."

https://www.cmu.edu/dietrich/news/news-stories/2017/october/...


It sounds like they used fingerprints correctly.

Wrong way to use fingerprints: "We have a crime and no suspects, but we have a fingerprint. We ran that through our database, and got a single match. That must be the person who left the print".

Right way to use fingerprints: "We have a crime and several suspects that we know were deeply involved in the matter that the crime was part of, but we don't know which were actually there during the crime. We also have this fingerprint left during the crime, and it matches exactly one of the suspects, which shows beyond a reasonably doubt he was there".

It's really no different from hair color, skin color, sex, height, weight, age, or DNA--the evidentiary value of a match on any of those depends on how that thing is distributed in the population that you can show by other means includes the correct person. The more you can narrow down that population, the better the evidentiary value.

Take sex for an extreme example. A woman reports that a man assaulted her in New York. There are a lot of men in New York, so if that's all you have to go on it is pretty worthless. But suppose an officer saw the attack, got a good enough look to see that the attacker is a man but nothing else, and saw that man run into a women's restroom to hide. The officer enters the restroom, and finds that it is occupied by two women and a man, and there is no way out of the restroom except the door the man entered by, which the officer had under observation continuously from the time the man entered to the time the officer entered and saw the three occupants.

Convicting that man for assault would be quite reasonable on the above evidence, even though it is largely based on the fact that he is male and the assailant was male. That's because other good evidence narrowed the population of possible assailants down to just three, and so if you can conclude that the assailant must be in that group, then the assailant being male does show that he was the assailant.


Fingerprints as used above wouldn't satisfy the "beyond a reasonable doubt threshold." They are circumstantial evidence of the identify of the suspect, but fingerprints aren't as unique as once claimed, nor are the imaging techniques (including resolution) referred to in the article sufficiently detailed for a proper fingerprint ID in the first case.

It's questionable whether the bathroom situation would satisfy the threshold, as it's based entirely on the testimony of an officer who somehow was able to ascertain gender but not (approximate) height, (approximate) weight or build, or race from his brief glimpse of the assailant. Moreover, it assumes the officer was correct about the gender of the assailant. Finally, it assumes that the officer was truthful in his statements. All in all, reasonable doubt exists at multiple levels and a good defense attorney only needs a single element of doubt.


They do. They don't care, because it helps the defense and not them.


>that while latent fingerprint examiners can successfully rule out most of the population from being the source of a latent fingerprint based on observed features, insufficient data exist to determine how fingerprint features really are unique

>>This makes it scientifically baseless to claim that an analysis has enabled examiners to narrow the pool of sources to a single person.

While it doesn't narrow it down to a *single person, Fingerprints are still evidence. The litmus is to convince a jury of your peers.


Ask the average jurist, “we found the suspect’s fingerprints at the scene, is it possible the suspect wasn’t there?” the answer 12/12 times will be no.

It’s even worse with DNA which could have been transferred.


> “we found the suspect’s fingerprints at the scene, is it possible the suspect wasn’t there?”

This is a perfect example of "begging the question"! Hopefully we can ammend the question with "we found a fingerprint at the scene that matches the suspect" and get that number down to 11/12.


Ask the average jurist, “we found the suspect’s fingerprints at the scene, is it possible the suspect wasn’t there?” the answer 12/12 times will be no.

A jurist is lawyer or otherwise an expert of law. You'd most likely get a panel of 10-12/12 jurists agreeing that it's possible the suspect wasn't there at the time the crime was committed.

You'd probably get a different answer if you asked a juror, but you still wouldn't get 12/12. Jurors are familiar with the concept of time. It's not even to prove that the suspect was there at all (especially if they have an explanation for why they were at the scene of the crime), the prosecution must establish they were specifically there at the time of the crime.


https://www.youtube.com/watch?v=XC2jpKfgKko

My guess is, they probably know but ignore it?


You could also have an evil twin brother that you don't know of that might have committed the crime, thus let's throw out photographic evidence.

Somebody might also put on a fake license plate number matching your own on an car with the exact model as your own. Thus let's throw out witness accounts of car type/plate number.


US Criminal law has a long long history of reliance on scientifically invalid forensic evidence. Bite marks, fingerprinting, fiber analysis, breathalyzers, hypnosis and others have all been legally ruled to be beyond reproach at various times/places, and all of them are on a continuum of scientific reliability ranging from laughable through potentially useful for some purposes if properly performed and evaluated.(Hypnotizing witnesses is still legally valid in Texas and some other places.)

If "beyond a reasonable doubt" is supposed to mean something, then it should mean something.

And by the way, eyewitness testimony is frequently really unreliable.


I don't think this is fair. The criticism of fingerprints, as I understand it, is (a) there's actually no way to prove that "no two people have the same fingerprints," and (b) the current system of fingerprint classification (using a few points) could allow different fingerprints to be classified as identical.


Sure, but folks using this argument don't quite use "evidence" in the same sense that courts do. Evidence is not "irrefutably correct", it is "a series of datapoints that judge (and possibly jury) need to evaluate as part of the decision."

What's important is to lay out accurate odds for these systems, to which the fingerprinting folks don't always do. But a 5 point match on a fingerprint IS information even if it's not wholly conclusive.


You are correct in a Bayesian sense, but until everyone is trained to be a Bayesian reasoner in primary school, we need to build the system in a way that ensures evidence is considered properly.


There is much ink spilled about how people aren't bayesian educated, but I don't really believe it and therefore don't agree with you.

We're actually pretty good at identifying uncertainty and certainty within the physical world. It's only when we get into very large or very small numbers that we start to lose the plot.


So what's the solution? Throw out all evidence that isn't incontrovertible?


Seems like you'd need to prove the pills in question were the alleged substance?

Just linking a picture of alleged narcotics to someone via their fingerprint doesn't seem like it should rise to the level of reasonable suspicion. I don't know what the standards are for the UK though, so I may be applying assumptions.


> Seems like you'd need to prove the pills in question were the alleged substance?

That may be true in order to convict him of dealing the ecstacy, but he wasn't. From the BBC article:

> It [the dealer's house] was raided and large quantities of 'gorilla glue' - a type of cannabis - was recovered.

http://www.bbc.co.uk/news/uk-wales-43711477

His sentence in the end was for conspiracy to supply cannabis. The suspicion to search was contextual, given that the image was obtained from another arrested person (presumably for a drug offence) along with other messages implying he was dealing to them, such as "what do you want to buy?".

Although I'm not sure how the fingerprint comes into this, given that they didn't find him by it, but rather "officers had an idea who they believed was behind the drugs operation", and he was convicted of a different crime, with evidence from his home search.


I don't think the pills matter. I think someone was caught with drugs, said "Here's the whatsapp conversation I had with the person that sent them to me" and received the fingerprint. I think they already had the drugs.


I seriously doubt that photo had enough of the fingerprint in it to get a match. I'm not saying this can't be done - I just don't see how that particular photo offers enough data points to uniquely identify someone.

But if anyone has more details on the technical analysis they did, do tell!


At how many points of the finger print they look is a whole different story and can vary widely.


Quick heads up to everyone who hasn't apparently read the article: This case is from the UK, where the evidentiary rules are different. It wouldn't pass muster in the US as there is an additional procedural hurdle the prosecution would have to clear to establish that the fingerprint(s) at issue could have been reliably extracted from the image. If the image in the article is the actual image in the case, it wouldn't satisfy US guidelines for fingerprint analysis.


Opens the possibility of framing someone by Photoshopping their fingerprint into a photo.


Interesting use of neural networks right there. Putting someone else's fingerprints on the hand of someone in a photo to make it look like they were the subject of the photo.


fingerprints and DNA are unique.... you can't fake/plant those /s


I think I'd be much more concerned about digital fingerprints that come from the lenses and sensor than physical finger prints. Anyone smart would hide revealing details. Sites who offer services like this should introduce noise to the images that distorts at least the digital artifacts.


Has anyone unlocked a phone using a picture of the owner's finger? That would be neat!


It's not as straightforward as using a printed image; fingerprint readers are designed to detect only fingers, but it is possible to fool them:

https://www.theverge.com/2016/7/21/12247370/police-fingerpri...




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: