Technology Review - Published By MIT
Technology Review  in English | en Español | auf Deutsch | in Italiano | 中文 | in India
Advertisement

Identifying Manipulated Images

Continued from page 1

By Erica Naone

Monday, March 17, 2008
smaller text tool iconmedium text tool iconlarger text tool icon

Neal Krawetz, who owns a computer consulting firm called Hacker Factor, presented his own image-analysis tools last month at the Black Hat 2008 conference in Washington, DC. Among his tools was one that looks for the light direction in an image. The tool focuses on an individual pixel and finds the lightest of the surrounding pixels. It assumes that light is coming from that direction, and it processes the image according to that assumption, color-coding it based on light sources. While the results are noisy, Krawetz says, they can be used to spot disparities in lighting. He says that his tool, which has not been peer-reviewed, is meant as an aid for average people who want to consider whether an image has been manipulated--for example, people curious about content that they find online.

Cynthia Baron, associate director of digital media programs at Northeastern University and author of a book on digital forensics, is familiar with both Krawetz's and Farid's work. She says that digital forensics is a new enough field of research that even the best tools are still some distance away from being helpful to a general user. In the meantime, she says, "it helps to be on the alert." Baron notes that, while sophisticated users could make fraudulent images that would evade detection by the available tools, many manipulations aren't very sophisticated. "It's amazing to me, some of the things that make their way onto the Web and that people believe are real," she says. "Many of the things that software can point out, you can see with the naked eye, but you don't notice it."

Johnson says that he sees a need for tools that a news agency, for example, could use to quickly perform a dozen basic checks on an image to look for fraud. While it might not catch all tampering, he says, such a tool would be an important step, and it could work "like an initial spam filter." As part of developing that type of tool, he says, work needs to be done on creating better interfaces for existing tools that would make them accessible to a general audience.

Tags

forensics photography software

Related Articles

Comments
  • Apollo
    I would be nice to use this tool to determine if any of the Apollo moon landing photos were faked.  Theoretically there should only be a single light source, shouldn't there?
    Rate this comment: 12345

    mdale9
    03/17/2008
    Posts:2
    Avg Rating:
    2/5
    • Re: Apollo
      Hi mdale9,

      There is one primary light source (the Sun), but there should be lots of ambient light of varying degrees (stars and reflection of any earthlight).

      As far as analysis goes, there is usually no difference between a "real" and "staged" photo -- both are "real" in that they were not digitally manipulated. However, this does not mean that the content is real. For example, Imad Mughniyeh staged the famous photo of him holding a gun to a pilot's head (http://edition.cnn.com/2008/WORLD/meast/02/13/btsc.clancy.mugniyah/index.html). The interpretation of the photo's content is deceptive but the image itself is real.

      When you ask about the moon landing, the photos that I have looked at appear to be real. However, I will not to get into a debate on whether they were staged or actually taken on the moon -- too many people ignore facts in place of agendas and conspiracies.

      - Dr. Neal Krawetz
      Rate this comment: 12345

      nealkrawetz
      03/17/2008
      Posts:3
      Avg Rating:
      4/5
    • Re: Apollo
      The surface of the Moon is lightly-colored and, with no atmosphere, subject to brilliant light from the sun - how can you say there's only one light source?
      Rate this comment: 12345

      smithsomian
      03/19/2008
      Posts:71
      Avg Rating:
      4/5
  • Identifying Manipulated Images
    I'm curious - is this only for digital photographs, or will it work with a film-based photograph that has been scanned and digitised?
    Rate this comment: 12345

    deejay
    03/17/2008
    Posts:23
    Avg Rating:
    4/5
    • Re: Identifying Manipulated Images
      My understanding is it would work with scanned and digitized photos as well as photos that were originally digital.

      -- Erica Naone
      Rate this comment: 12345

      Erica Naone
      03/17/2008
      Posts:49
      Avg Rating:
      4/5
    • Re: Identifying Manipulated Images
      Hi deejay,

      It depends on the type of manipulation. Light direction may still be detectable, as will any poorly pasted borders. However, my tools also look for compression artifacts and format anomalies. These will not be present if a photo is manipulated before being digitized.

      - Dr. Neal Krawetz
      Rate this comment: 12345

      nealkrawetz
      03/17/2008
      Posts:3
      Avg Rating:
      4/5
  • New moon & mars pictures
    Is this tool available somewhere, or is it source closed ? I think it would be nice to test the images on the new moon images (specially those from the chinese probes) and the mars pictures (which are said to be [color-]tampered, although maybe it wouldn't work for color tampering detection).
    Rate this comment: 12345

    dsouza
    03/17/2008
    Posts:1
    Avg Rating:
    2/5
    • Re: New moon & mars pictures
      At this point, neither of the tools mentioned in the article are publicly available, as far as I'm aware. I should also note that, while other digital forensics tools might detect color changes, tools that pay attention to lighting (which I've focused on here) seem best suited to flagging cases where multiple photos have been combined into one.

      -- Erica Naone
      Rate this comment: 12345

      Erica Naone
      03/18/2008
      Posts:49
      Avg Rating:
      4/5
  • A picture can be a thousand lies.
    Even if such software tools are available, one could still produce fake images.

    As a forger, you would take the fake image and run tests on it with the forensics software, and see which part of the image is flagged as fake. Then you modify the image accordingly, and test again. A few iterative steps like these, and you would be able to produce an image that passes the test.

    Sure this means more work on the part of the forgers, but then they can claim to have a "genuine" picture.

    We should never rely solely on pictures as proof anymore. A picture can tell a thousand lies.
    Rate this comment: 12345

    gabrielg01
    03/18/2008
    Posts:418
    Avg Rating:
    3/5
    • Re: A picture can be a thousand lies.
      You're absolutely right. However, when I asked Johnson about this, he pointed out that, for now, forgers tend not to have access to these tools. It'll be a different game once they're packaged with Photoshop.

      One additional point is that, like all questions of security, if someone is intent on pulling off a beautiful and sophisticated forgery, and has the skills to back that up, they probably can. However, many forgeries simply aren't done with that level of dedication and skill, and tools can help filter out those forgeries.
      Rate this comment: 12345

      Erica Naone
      03/19/2008
      Posts:49
      Avg Rating:
      4/5
      • Re: A picture can be a thousand lies.
        One of the things I mentioned at my Black Hat Briefings talk was specifically about the necessary skills. The tools are widely available -- Photoshop, Gimp, 3DS Max, Maya, Bryce, etc. either come with your system (or camera or scanner) or can be purchased as off the shelf software.

        Having the tools is not difficult. However, having the skills to use the tool is a different story. (Just because I own a hammer does not mean I can build a house.) Most people who modify images lack the skills to overcome most image analysis techniques. In fact, many forgeries by professional artists still test positive for being modified.

        My own tools are around 86% accurate. The few misses were by professional artists who have won awards from the CG Society. And two of the misses were by the same artist who spent over two years on the same model (wireframe, skin texture, hair, etc.). Fortunately for detecting forgeries, most amateurs don't have the skill or time to devote to their edits. (Even I, knowing exactly how the tools work, cannot make a good forgery because I lack the skills to use the drawing tools at a professional level.)

        Given enough time, an amateur artist should be able to defeat a few of the detection methods. However, I use a suite of analysis tools and methods -- light analysis is just one approach. Very few doctored images can make it past all of the tests, and those that do were likely by professional artists who have plenty of time to devote to the forgery.

        You are correct that a professional forger could use these forensic techniques to better the forgery. Fortunately, most doctored images are not by professionals, and even professionals may not have the time to overcome all of the signs of modification.

        There is a saying in the computer security field: the defender has an impossible task; they must defend against all possible attacks, while the attacker only has to find one vulnerability. The same can be said about photo forensics: the forger must hide all possible signs of modification or forgery, while the forensic examiner only needs to find one clue.
        Rate this comment: 12345

        nealkrawetz
        03/19/2008
        Posts:3
        Avg Rating:
        4/5
  • Identifying Fakes
    Now if we could figure out a way to tell when Hillary is lying; oh yeah thats right we did, when her lips are moving.
    Rate this comment: 12345

    jmaximus9
    04/04/2008
    Posts:85
    Avg Rating:
    3/5
  • Limitations
    "Different manipulations will be spotted by different tools"

    beware the false positive...

    ----------

    the other consideration here is that hollywood has long ago mastered the technique of matching the lighting of separately filmed objects.  I doubt any of these tools will succeed against modern techniques of movie making.  ILM etc are amazing in what they do.


    ---------------

    as far as the moon fake goes...   will people pleaazzzee  get a clue.  First you must consider the technology of the time.  Certainly today hollywood could fake it fairly convincingly.  But back then nobody had any equipment that even came close to being able to do what is proposed, those cameras where primitive.  Then consider the size of the stage that would have been needed for the rover to run around in.  And in a total vacuum... even today it would be exceedingly difficult to build an open chamber that large and keep it from collapsing.  14 pounds per square inch does not sound like very much, until you try to stop it from crushing something large.

    And finally consider gravity which everybody suggesting these ludicrous ideas always ignores. You simply can not fake the way the astronauts  were bouncing around up there (both on foot and in the rover).  You can not fake the speed with which the dust settled and the hammer fell.  No I am not talking about the fact of it being in a vacuum, I'm talking about the rate of fall which is determined solely by gravity.

    And on top of all of that try *thinking* about how long it would take and how much effort it would take to create a movie like that.  It takes hollywood years of effort with vast resources to create their special effects films.  Far easier and faster just to go there for real...

    Try taking Physics 101 and stop buying$ this total krap about the moon landing conspiracy -- it's a great money $ maker for certain people.  The main conspiracy argument is that it was terribly difficult to do therefore we did not do it but faked it instead.  Yes it was difficult maybe even impossible, and that's what made it such a great achievement for good ol yankee ingenuity.  Remember that self-appointed ~expert~ of the year 2000 scare?  The guy didn't even know how to write a computer program, he had no technical background at all, but his "chicken little" routine made him a lot of money and generated a lot of hysteria, none of which came true.  There are a lot of these ~people~ running around. I suppose the Wright brothers faked it too since everybody knows that powered flight is ~impossible~.

    I recently saw a documentary on the building of the Alaska Canada Highway, and guess what, that was impossible too!!!   But somehow we did it anyway.  You want to see blow by blow of how people often do the "impossible", go watch that documentary, it's from PBS and probably at your library.


    Rate this comment: 12345

    smithsomian
    02/13/2009
    Posts:71
    Avg Rating:
    4/5
Reply

Log In

Forgot your password?     Register »
Advertisement

Videos
Latest Videos
Malleable Maps, Artistic Robots and Bubble Interfaces

Top Stories Of The Day

Technology Review January/February 2010

Current Issue

Security in the Ether
Information technology's next grand challenge will be to secure the cloud--and prove we can trust it.
Advertisement
Advertisement

RSS Feeds

xml icon TR Top Stories
xml icon TR Editors' Blog
xml icon TR Video Blog
xml icon TR Video
Advertisement
Subscribe to Technology Review's daily e-mail update. Enter your e-mail address

TECHNOLOGY RESOURCES
Advertisement
MIT Massachusetts Institute of Technology © 2010 Technology Review. All Rights Reserved.