Follow Slashdot blog updates by subscribing to our blog RSS feed


Forgot your password?
Biotech Crime Science

Police Use DNA To Generate a Suspect's Face 100

An anonymous reader writes: The NY Times has a pair of articles about a technology now being used in police investigations: computer generation of a suspect's face from only their DNA. Law enforcement in South Carolina had no pictures or descriptions of a man who murdered a mother and her daughter, but they had some of his DNA. From this, a company named Parabon NanoLabs used a technique called DNA phenotyping to create a rough portrait of the suspect's facial features, which the police then shared with the public.

The accuracy of these portraits is still an area of hot debate — most of them look rather generic. The NY Times staff tested it with a couple of their employees, circulating the DNA-inspired portraits and seeing if people could guess who it was supposed to be. None of the ~50 employees were able to identify reporter John Markoff, and only about 10 were able to identify video journalist Catherine Spangler. But even though the accuracy for a person's entire face is low, techniques for specific attributes, like eye color, have improved greatly. Of course, the whole situation raises a slew of civil liberties questions: "What traits are off limits? Should the authorities be able to test whether a suspect has a medical condition or is prone to violence should such testing be possible?"
This discussion has been archived. No new comments can be posted.

Police Use DNA To Generate a Suspect's Face

Comments Filter:
  • by Nidi62 ( 1525137 ) on Tuesday February 24, 2015 @09:21AM (#49118283)

    "What traits are off limits? Should the authorities be able to test whether a prone to violence should such testing be possible?"

    Because it's hard to tell if someone who killed 2 people is violent....

    • by Anonymous Coward

      But they will try to use the genetic evidence you might be violent in some circumstances to prove you did a specific murder. Its basically racial profiling to the max

    • Re: (Score:2, Insightful)

      by Anonymous Coward

      The most murderous people sit behind podiums. The next most murderous sit in labs. Then you have those who quietly and with great discipline aim the weapon or prepare the poison.

      The least dangerous murderers are those who lost their shit in the heat of the moment. Which isn't to say they're not dangerous, but there are a lot of people who could reach this state of mind, given sufficient provocation. Perhaps most people?

    • by oodaloop ( 1229816 ) on Tuesday February 24, 2015 @09:52AM (#49118439)
      I think the point is if they take a suspect's DNA and show that he's prone to violence, and use that as evidence he committed the crime, rather than taking DNA from a violent crime scene.
  • Remember when we worried that perverting science was being done in the name of some shark killing?
  • Suspect (Score:2, Funny)

    by Anonymous Coward
    The suspect is described as slim, 5' tall, and double-helical.
  • Real helpful (Score:1, Insightful)

    by Anonymous Coward

    DNA says that the suspect is an African-American male, with brown eyes, dark hair, and dark skin. Anyone with any information on anyone matching the description, please contact the authorities.

    • Re:Real helpful (Score:5, Interesting)

      by penguinoid ( 724646 ) on Tuesday February 24, 2015 @10:12AM (#49118541) Homepage Journal

      Even very weak evidence is useful, even if it would be too weak for court. If you know* the perp is African-American, you can't go around suspecting everyone who's African American, but you most certainly can eliminate all your white/asian/hispanic suspects.

      *Sadly/amusingly, eyewitness accounts are not sufficient for this.

      • by DM9290 ( 797337 )

        Even very weak evidence is useful, even if it would be too weak for court. If you know* the perp is African-American, you can't go around suspecting everyone who's African American, but you most certainly can eliminate all your white/asian/hispanic suspects.

        *Sadly/amusingly, eyewitness accounts are not sufficient for this.

        not necessarily. you will be shocked to learn that race is not a phenotype.

  • That's great! (Score:4, Insightful)

    by hcs_$reboot ( 1536101 ) on Tuesday February 24, 2015 @09:41AM (#49118389)
    When that technology will have evolved enough, it'll be able to show what you were supposed to be from your DNA, compared to what you actually are. How your body and face have changed due to your family, education, school, company etc... That technology will contribute to settle the debate "genetics vs environment".
  • ...However, if any of those other countries did the same thing, it wouldn't be OK. In fact, we would be thanking God for the USA and its guarantee of "freedom & liberty" - the Snowden saga not withstanding...

  • by gl4ss ( 559668 ) on Tuesday February 24, 2015 @09:46AM (#49118413) Homepage Journal

    I'm pretty sure it's the lead character from GTA:SA.

    seriously, it's pretty friggin generic. the company does good money probably on it.

    how about.. along with ordering this from them.. order known faces+dna pairs and see if it's any good.

    • how about.. along with ordering this from them.. order known faces+dna pairs and see if it's any good.

      I know, I know... it's still unfashionable to RTFA. But I did it anyway, so you won't have to:
      They did exactly that in a related article linked from there: []

      At least a dozen people immediately responded that they could not guess because the images felt too generic. Among the 50 or so people who did venture guesses, none identified the man as Mr. Markoff, who is 65.

      When it ca

      • by phorm ( 591458 )

        Yeah, just looking at the "real" vs "generated" photos.
        Noses don't really match up on most of them, nor do eyebrows. The chin is sorta close but without any of the other defining facial characteristics (age lines etc) it's pretty ambiguous. Also, the generated photos have cut out most of the upper-jawbone area so one can't even compare that.

        IMHO they seem to be able to give you a decent approximation of the skin tone - minus blemishes - but not much other than that. I'd think that these might actually be co

      • by gl4ss ( 559668 )

        the article with real vs. made pictures is about the method by some university person, while the police are using a company.

        the method MIGHT be the exact same, but is doubtful.

        basically what the genome will tell is just ethnicity(color, nose shape etc sterotype stuff really). basically what I'm getting at is that a VERBAL description made from the same information could be more useful - and yeah I'm pretty sure the coppers in case knew already that the dna was a "black male". now we have a sims(1) quality p

  • by Anonymous Coward

    It was either this guy [] or this guy [].

  • by Sparrowhawk7 ( 1361853 ) on Tuesday February 24, 2015 @09:46AM (#49118417)
    Sounds more like a case of Parallel_construction [] to me. Its not like the technique is unknown to law enforcement. With all the additional pressure around Stingray (cell site simulator) use, I fully expect these types of techniques to proliferate.
  • This is creepy! (Score:4, Interesting)

    by Aethedor ( 973725 ) on Tuesday February 24, 2015 @09:53AM (#49118445) Homepage

    Why is privacy so important? Because you don't know what creepy things governments will do with it in the future. All the condition under which you gave away some of your personal information might not apply in the future. And getting your information back at that time will very likely be no option.

    What if your face ends up with this new creepy technology. How can you even possibly defend your self against it? Some, for normal people, impossible to comprehent scientific research apoints you as a suspect. What can you do? This is creepy and scary and not something we should want.

    • Re:This is creepy! (Score:5, Interesting)

      by Dr. Spork ( 142693 ) on Tuesday February 24, 2015 @10:27AM (#49118631)
      I think we'd feel much better about it if we used this tool to remove suspicion from people, rather than to add suspicion. For example, this tool could probably rule out that the suspect is black or asian, that it's not a woman, it's not someone over 5'9, etc. Using the tool to generate a crappy portrait is the real bad move, because if you look like that, people will think that's evidence for your guilt. If this tool were only used to exonerate people and to remove them from the suspect list, who could object?
    • by Anonymous Coward
      My parents have already met my doppleganger. He lives in Finland, so there's little chance of a mistake in identity, but he was so similar that they did a double take. They showed me a picture and I thought it was me. Very disconcerting. If someone else's DNA can produce your face, or worse, the computer produces your face from someone's DNA but no one else in the world actually looks like you, the you're screwed.
  • by nimbius ( 983462 ) on Tuesday February 24, 2015 @09:56AM (#49118469) Homepage
    The most recent use of this technology was in tracking down the Baton Rouge serial killer [], although the same problem exists here as does DNA evidence. Namely that close enough is good enough [] in the eyes of a court of law, particularly in the southern half of the United States.

    There was some argument that Derrick Lee was perhaps incompetent to stand trial; during psychiatric evaluations he scored an average of 65 on various standardized I.Q. tests, and a score below 69 is considered to be the threshold for what can be considered mental retardation. Lee was, however, deemed fit to stand trial.

    But like phrenology, lie detectors, and to some degree the shrouded witchcraft code of the breathalizer, its a modern tool in the fight against "the bad guy"

  • by dingleberrie ( 545813 ) on Tuesday February 24, 2015 @10:00AM (#49118491)

    Because it searches for correlation between faces and DNA, and it's getting arguably discernible results already. I'm expecting it to improve as it gets more sample data, more processing power, and more researchers identifying distinguishing facial characteristics for it to attempt a DNA correlation to. Further, when they find out how to show examples at different milestone ages, then that would lead to even more interesting applications. Imagine knowing what your baby will look like before they are born... and the societal questions that that brings.

  • I have no problem identifying these perps. [] []

  • Snake oil (Score:2, Interesting)

    by Anonymous Coward

    Sounds a lot like snake oil to me. They have cloned domestic cats and found that the identical DNA used in the cloning produces totally different fur patterns. Here's an explanation why [[]]. While human beings are not cats, it explains that when cells differentiate, random groups of cells become different things (legs, lungs, heart, etc.). Some things (like eye color) may be fixed, but everything else is negotiable. External considerations can also appl

    • by Anonymous Coward

      It's not comparable, at all. Fur pigment is one thing, morphology is another entirely. And we know this because two people with the same DNA are indistinguishable from one another by most people. We call them identical twins.

      • by Anonymous Coward

        Identical twins gestate under identical circumstances, in the same womb given the same nutrients/chemicals and are more than likely after birth raised in a very similar fashion. If you somehow snatched one of the eggs and put in in another womb of a woman living in different circumstances I think you'd be surprised how much of a difference there would be.

      • A good example, and one might also note that identical twins have different fingerprints (for the same reason cloned cats have different fur patterns).

    • It may be snake oil now, but it sounds like it has a lot of potential. There are a lot of features that are genetically determined, or genetically determined within constraints. It seems to me that generating a computer-generated face is a terrible way to use this, as it seems like it would be misleading in both the way that might implicate an innocent, yet also excuse the actual perp (see also how most people didn't recognize a person from their generated face). To use this properly would probably require

      • The company is hiring a Computer Vision Engineer (and several others) for this product line, so I'd say they have plans for significant improvement down the road. Things like dynamic aging, expression rigging, rendering multiple facial variations to highlight environmental variability to the genetic plan (I expect a lot more data will need to be gathered before this in particular is implemented), perhaps even computer-based recognition as suggested... Snapshot can definitely be misused as other commenters h
      • I take back what I said about implicating an innocent -- being a DNA based approach, this won't implicate an innocent past the point where they ask for a DNA sample to compare directly with the DNA sample they used to generate the image.

  • I can't wait to see the result:
    a) They catch the culprit and he looks exactly as the DNA said he would.
    b) They catch the culprit and he doesn't look as the DNA said he would.
    c) Either a) or b) plus the one they catch is actually innocent.

    How they tell which is the truth, who knows?!

  • ...I'd just be happy if our local news stations would share with us such basic facts as the skin color of the bloody suspect.

    They seem to avoid it unless the criminals are white or asian, for reasons that likely depend on your political bias; they are failing to describe brown-skinned criminals ....
    a) because they're conservative, and assume that you assume 'criminals are brown anyway', or
    b) because they're liberal, and they don't want to confirm the stereotype that criminals are mostly brown

  • My understanding from what I have read about genetics is that usually genetics only affect about 50% of what makes you, you. The rest of the 50% is due to environmental conditions. A mug shot of you that's only 50% accurate is going to be a challenge to use properly.

  • by Anonymous Coward

    Often the police can't catch anyone unless someone rats out
    the perpetrator.

    This stuff is just another way to conceal the true idiocy of the police,
    who are pretty much useless unless you are in the government and
    need your opposition to be suppressed.

  • If criminals know their DNA will get a thorough health checkup while they can't personally afford such tests....

    Some may feel that committing crimes and leaving DNA samples is the only way to get such quality information.

    So then the police, in an effort to stop such medically motivated crimes decide to withhold the information which does not directly relate to identification... Leading said criminals to file Lawsuits demanding their own medical data.

    Orrrr... The police just run it through the secret DNA da

  • "None of the ~50 employees were able to identify reporter John Markoff, and only about 10 were able to identify video journalist Catherine Spangler."

    So, a crime has been committed, there are ~50 witnesses, of which 'only about 10' are able to identify one person. Statistically significant? What would the police and the courts think?

    Oh, and what exactly is 'about 10' people? Somewhere north of 9.75?
  • ... it would be accurate for someone named 'Dan' who is dyslexic
  • Awesome! How long before I can get a lock of hair from someone, then auto-generate an avatar of them?

  • They might as well generate his dental work, or diagnosis and cross reference the likely cancer treatments he is likely to have with purchases of any vitamin supplements he may require as a result of genetic deficiencies.

    Whats that.. four datasets.... Credit Card History, Dental History, Medical History, Facebook photo. Identify likely diagnoses and filter. Obviously the older you are the easier it is to trace you since you are likely to have certain genetic risk become issues. A few good generic markers an

  • No, DNA phrenology... Read the bumps in the helix. That shit still isn't dead yet.

  • reminds me of a tv documentary, I think of an investigation in Ireland. The perps had left an apple at the scene that had some unusual bite marks in it. The Dentist they showed it to gave a full description of the guy, down to hair color, how tall he was. Apparently some rages/genders/body types were genetically predisposed to gnawing on apples in that way. Of course, now I know about parallel construction... But it made for a great tv show.
  • by allo ( 1728082 ) on Tuesday February 24, 2015 @12:40PM (#49119679)

    Sounds like a great idea.

  • The only thing that comes to my mind is this quote:
    "his face is so generic it matches every other face in our database" []
  • We have generated your crime profile. Your arrest will be timed to 3 months prior to your crime.

    Move along, citizen.
  • Looking at the images, it seems that the discrepancies are mostly related to lack of skin complexion details (exact color, texture, sheen), all the sorts of "minor" details that kick many CGI rendered human images into Uncanny Valley []. The object rendered looks human-like but our minds scream NOT HUMAN when we don't detect these minor details and cues.
  • Wasn't there a dramatic murder gene failed defense in the 90s about people genetically programmed to murder? What if DNA analysis found this to be very true? What if you had certain gene patterns that matched historical prisoners that were found guilty of first degree murder? You know what I would do if we lived in a world like that? Learn how to obfuscate, change those DNA patterns. Then the decision is, do I give it away to everyone or only people that I trust would not murder someone. That power would b
  • They could release a facial construction along with a list of heritable traits such as if the perpetrator has cystic fibrosis, Crohn's disease, lactose intolerance, sickle cell anaemia, or a number of other traits that can be tested for using DNA. If the person tested positive for a rare disease it could go a long way towards finding them.

I have ways of making money that you know nothing of. -- John D. Rockefeller