Forensic Photo Analysis

Questions, speculations & updates on the techniques and nature of media fakery
Flabbergasted
Administrator
Posts: 1243
Joined: Mon Nov 12, 2012 12:19 am

Re: Forensic Photo Analysis

Unread post by Flabbergasted »

Flabbergasted wrote:
nonhocapito wrote:Another possibly interesting read on the subject:

Adobe Photoshop Forensics
Sleuths, Truths, and Fauxtography


Image
from http://www.amazon.com/Adobe-Photoshop-F ... 1598634054

...this one can be found in electronic format via, er, many channels. :rolleyes: .
nonhocapito,
I tried to find one of those channels, but no luck so far. Any way you could give me a hand?

Edit: Just had a lucky strike ... no need any longer! :)
I just had a closer look at this book. It does contain some good sections and observations, like this:
The Neverending Fraud
Outside the scientific community, we don’t always realize how serious such fakery can be. But bad science contributes to the misuse of millions of dollars in government and corporate grants. It misrepresents reality to the public, creating panic or prompting bad political and social decisions. It can delay medical cures by misdirecting effort and funding to fantasyland.
And once a bad paper gets published, it lives a kind of half-life in the community. Even if the authors retract a paper (or are asked to retract it, which is much the same thing but considerably more embarrassing), the paper is still searchable online and continues to live in its original form in libraries and labs. Doctoral students looking for citations to bolster their own work may continue to cite it.
But, not surprisingly, the book downplays the massive proportions of media fakery over the past century by not providing any examples with truly damning implications. For example, in a small section on NASA´s moon pictures, the only issues mentioned (not really discussed) are "light reflection from the lunar surface" (to explain away clearly lit objects in the shade) and "visual clues and angles of view of distant objects". The author, whose credentials include...
Associate Director of the graduate program in Digital Media and the undergraduate program in Multimedia Studies at Northeastern University in Boston, MA. Formerly she was the Technical Director for the Department of Visual Arts, and she has been teaching courses that integrate digital tools with design and creative imaging for 20 years.
...apparently found no examples of fraud worthy of her attention in the entire collection of moon imagery. Here is the full text of the section (highlights added):
One Small Step
If you were beyond preschool age in 1969, chances are you remember a summer night in front of the TV, as Neil Armstrong and Buzz Aldrin walked on the surface of the moon. It was the culmination of several days of anticipation and excitement that began when Apollo 11 was launched from the Kennedy Space Center in Florida on July 16 (Figure 9-2).
Surprise! As much as 10 percent of the American public believes that the event never took place—that it was an elaborate hoax supported by falsifying everything from sound transmissions to rock samples. The argument: that the flight and landing were technologically beyond NASA’s capabilities. Recognizing the impossibility, the government took the money earmarked for the moon flight and spent it on faking the event and bribing people into complicity. The motive: Cold War politics and national prestige.
There are several variations on the Apollo hoax theory. They range from the absolutist—everything about the event was faked—to the hybrid theory—the capsule was shot into space but didn’t go all the way to the moon, and the astronauts taped the landing in Nevada before the flight. Although the theorists bring up everything from missing telemetry tapes to moon temperatures and radiation, the published pictures from the landing are the most frequently cited evidence.
Is there any support for these accusations? Well, not based on the photographs—the arguments lack a photographer’s understanding of light. For example, most of the bullet points for a faked landing claim that there would be only one source of light on the moon. If so, an astronaut standing in a shadow should be totally dark. Since you can see the astronauts clearly when standing in the shadow of the lunar module, there must have been a second light source.
They’re right—there is a second light source: reflected light from the first one. Think of being on the beach on a sizzling hot summer day, and how much brighter and harsher everything appears because of the reflection from the sand. The lunar surface would reflect light as well, with no atmosphere to dampen the effect. Note that reflected light is the only way to explain the secondary shadows on the backpack, inner arm, and lower legs of Buzz Aldrin in Figure 9-3.
Another claim is that NASA used a backdrop of mountains, and then they erred and took a picture without the lunar module in place (Figure 9-4)[the lunar lander is not even shown in this figure]. But two things stand in the way of this idea. The first is that there is no lunar atmosphere, so we don’t have the same visual clues of distance that we do on Earth, where the atmosphere makes distant objects appear fuzzy and blue. The second is a standard of observation that anyone who has ever driven down a flat road in the desert has seen. You have to go quite some distance toward a far-away object before it feels like you’ve made any progress toward it.
After 30 years and many administrative and political changes, not to mention several follow-up landings tracked around the world, scientists of every political stripe agree that the landing, and all the moon visits, really did take place. But as long as the photographs exist and the element of mistrust remains, some people will disbelieve.
The short section whitewashing NASA´s moon pics is followed by a quite long section debunking UFOlogy and urban legends. Perfect. Let´s get all the kooky theories together in one place.

Another section provides...
a lovely example of how a real image takes on a hint of the faux, without the edit of a single glorious pixel.
Image

The book shows how several images of the "extremely photogenic" Helix Nebula are combined to make the picture we see in school books and science magazines. According to Baron, it´s...
...a beautiful illustration that’s a tribute to the marriage of art and science.
Subsequently, a single example of a fake 9/11 picure is given: The silly "tourist guy" hoax!

Image

Anything else we should know about, Ms. Baron?

Nope...

Image
Flabbergasted
Administrator
Posts: 1243
Joined: Mon Nov 12, 2012 12:19 am

Re: Forensic Photo Analysis

Unread post by Flabbergasted »

Meet Cynthia L. Baron (MBA in Marketing) from Northeastern University, an acknowledged connoisseur in photoshop forensics, sleuthing, truthing and moon fauxtography:

Image
Cynthia L. Baron is a faculty member and the academic director of the Master of Professional Studies in Digital Media program. She is responsible for curriculum and program development, as well as adjunct faculty recruitment, scheduling, and supervision. Prior to that, Ms. Baron was the associate director of the Multimedia Studies program and served as technical director and clinical lecturer in the Department of Art & Design.

Her most recent publications include Designing a Digital Portfolio, now in its second edition, and Adobe Photoshop Forensics: Sleuths, Truths, and Fauxtography. She has been profiled or quoted in various media, ranging from MIT Technology Review to USA Today.

Previously, Ms. Baron was the cofounder and executive vice president of the graphic design firm Serif & Sans, Inc. She also was a series editor for Rockport Publishers, as well as a contributing editor for Critique and Computer Graphics World.
http://www.cps.neu.edu/faculty/faculty- ... -baron.php
anonjedi2
Member
Posts: 860
Joined: Mon Dec 31, 2012 5:50 am

Re: Forensic Photo Analysis

Unread post by anonjedi2 »

The Fotoforensics website was a victim of a DDoS attack on Friday, hours after exposing a fake photo from Syria. Interesting details on the blog.

http://www.hackerfactor.com/blog/

and more here:

https://plus.google.com/100685145195323 ... q5wd4e9kkZ
Sukiari
Member
Posts: 44
Joined: Mon Oct 14, 2013 11:06 pm

Re: The Ukrainian theater

Unread post by Sukiari »

kickstones wrote:
brianv wrote:Image

Madame Tussaud's Putin. The dummy on the right was pasted in from another image, while the dummy on the left is just a dummy, period.
Image created with the Linux GD library. All other EXIF stripped.


Image
Brian, this 'event' reminds me of The Hacker Factor exposure last year (see below). It showed how the Kremlin manipulated images.
Feeling Hungary

A few days ago, I saw an interesting political picture at FotoForensics. It shows Putin (right) and Viktor Orban (Prime Minister of Hungary, left) with Hungarian President Janos Ader in the middle signing papers. The picture caught my attention because of the frame over the mantel. It looks like it should be a mirror, but it isn't reflecting anything.

Image

The picture was uploaded to the Hungary news site "index.hu" on 11-Feb-2014. However, I couldn't find the article that went with it. (I blame my inability to read Hungarian.) What I did find was far more interesting. I sent the picture to TinEye and immediately found the source picture... sort of:

Image

This picture shows Putin and Orban in the exact same pose. It is the same camera angle, their clothing has not moved, their fingers have not moved... This is the same picture, but the man in the middle has been replaced by flowers on the table. The flowers picture was released on 31-January-2013. Back in January 2013, Orban met with Putin to discuss a variety of topics, including energy. A year later, on 15-Jan-2014, they met again and signed an agreement to expand a nuclear power plant.

There are more problems with this picture than just the digital addition of the man in the middle. For example, PCA measures JPEG actifacts. Look for the amount of detail (or lack of detail) within each JPEG grid. Visible JPEG grids indicate a low quality picture. Large blocks within each grid is a medium quality, and fine pixelated detail is high quality. (Ignore the coloring, look at the detail. The coloring is just to help highlight details, like dropping dye onto a microscope slide.) In this case, the quality seen on the man in the middle is different from the rest of the room -- so he was added. The lower corners of the picture frame/mirror are also different, so they were modified.

Image

At this point, we can be certain that the 2014 picture is a digital composition. However, the biggest clue appears when we apply a basic color histogram to the picture:

Image

The histogram shows a distinct white outline between Putin, Orban, and the background. The entire picture looks like a digital composition.
Brian, going back to the image you posted, if you enhance the below imge by x6, white outlines are also visible surrounding the three characters in the image.


Image

As The Hacker Factor points out:

'Media outlets use pictures to help convey a story. But if the picture that supports the story is fictional, then what does it say about the actual event?'


http://www.hackerfactor.com/blog/index. ... s/P13.html

http://eng.kremlin.ru/news/23569
While I don't post all that often on this site I really appreciate it and all the hard work that goes into many of the posts here. While I agree with the premise that there's a lot of fakery out there, we have to be careful with drawing conclusions based upon image analyses of low-resolution, highly compressed web-sized images. These images are going to be reduced in size, have very powerful sharpening filters (essentially edge enhancement), and the information density in solid-hue and nearly solid-hue background items like the sky, streets, and walls which lack much detail is going to be very low.

I don't think that it is useful to attempt to prove anything with techniques involving Photoshop or other filtering programs UNLESS one can describe, and then demonstrate (and falsify!) the premise one is working under. In the case of digital image manipulation, we are talking about mathematical equations, so ideally a person can describe their fakery-detectinon efforts in a mathematical way.

When there are very high quality, uncompressed images available (such as NASA's moon landing images) I think it becomes much easier to detect photo manipulation with the use of filters. However, when the images you are starting with have already been munged before you even see them, you have a much tougher time detecting anything except the compression artifacts and image quality filtering that is applied when the image is sized and compressed for web publication. A useful experiment for anybody who wants to understand this is to simply do it themselves. Acquire a RAW file from a modern digital camera, and experiment with the filtering techniques used to detect manipulation. Then resize the image so it is much smaller, enhance contrast, sharpen, and finally compress the heck out of the image and try your same technique. You will notice that the image now has a multitude of artifacts that it lacked before. Then mentally ask yourself: "Are these artifacts similar to the ones that are supposed to prove that an image was manipulated?"

Some image manipulations are very obvious. Cloning of a small crowd into a large one has been pointed out before, and this becomes undeniable even when the image fakers use transforms to attempt to change the basic geometric nature of the cloned image portions. I suspect, though, that in many cases the manipulation is not digital, but rather set-ups are portrayed as real events. There are many posts on this site pointing out obviously *staged* photographs, as well as other analyses of incongruent shadows and other excellent evidence that the photo was not taken where it was intended to depict. A good recent example of this is the Japanese hostages, who have converging facial shadows and the spotlessly clean captor who is supposed to be in the middle of a dusty desert landscape.

When a person simply shows a filtered version of an image that was already manipulated (sharpened, contrast, color, etc) and then compressed to hell, especially in a context-free post, this just serves to lengthen the page which is not helpful. As another person pointed out, this meeting is portrayed as really happening, including video, interviews, jets and motorcades (supposedly) and the whole nine yards. Would it be necessary to photo-composite an image of this meeting? Could one analyze different videos and photos of the event and come up with similar signs of manipulation? I fear that attempts to use digital techniques for detecting photo-manipulation are not reliable especially without some kind of mathematical or programmatic theory backing them up.
hoi.polloi
Member
Posts: 5060
Joined: Sun Nov 14, 2010 7:24 pm

Re: Forensic Photo Analysis

Unread post by hoi.polloi »

While I strongly disagree with the presumption that leaders do not Photoshop seemingly casual pictures (and indeed, the basic necessity of a picture to appear casual is what supposedly — in fact, these days, rather weakly— argues for its credibility), I agree with fbenario that you have brought an important topic up in the Derailing Room and I am moving it here. Feel free to stuff it back in case you find it distracting, fellow mods, but this is an appropriate topic heading under which to debate digital forensics.

Sukiari, your idea and contention is that some legitimate images should display signs of digital manipulation. (In this case, I wonder, what is the reason for the news deliberately camouflaging totally faked ones with such supposedly "legit" image practices in our days of incredible Hollywood special effects, where distinguishing is important?) Simon has made this point before, and particularly the point with the forensics tools. He seems to be frustrated primarily with the mere controversy around the tool, even though it has been shown to be useful in some cases.

I think the place we are at now with it, in case you or others were wondering, is that it is good to check. It is a relatively harmless tool to use to see if something weird comes up. While not the most powerful tool we have (which I would argue may be our own "common sense" to borrow an obnoxious but appropriate turn of speech) it is a tool. And when it is used properly, it can make the possibility of fakery compelling.

Perhaps it is time we requested specific guidelines on how to use it to prove fakery, even though this may discourage casual people from filling a thread with potentially useful information; the benefit could be that it would only do so when someone has actually found something. We could consider adding this to our required reading at the start of someone's registration.

Also, just in case you didn't really read the whole discussion that was moved there, you may have missed a funny little thing; 2013 and 2014 pictures of the Putin meeting, overlaid by Simon, results in this:

Image

Still think the event was as real as depicted in the media?
Sukiari
Member
Posts: 44
Joined: Mon Oct 14, 2013 11:06 pm

Re: Forensic Photo Analysis

Unread post by Sukiari »

hoi.polloi wrote:While I strongly disagree with the presumption that leaders do not Photoshop seemingly casual pictures (and indeed, the basic necessity of a picture to appear casual is what supposedly — in fact, these days, rather weakly— argues for its credibility), I agree with fbenario that you have brought an important topic up in the Derailing Room and I am moving it here. Feel free to stuff it back in case you find it distracting, fellow mods, but this is an appropriate topic heading under which to debate digital forensics.

Sukiari, your idea and contention is that some legitimate images should display signs of digital manipulation. (In this case, I wonder, what is the reason for the news deliberately camouflaging totally faked ones with such supposedly "legit" image practices in our days of incredible Hollywood special effects, where distinguishing is important?) Simon has made this point before, and particularly the point with the forensics tools. He seems to be frustrated primarily with the mere controversy around the tool, even though it has been shown to be useful in some cases.

I think the place we are at now with it, in case you or others were wondering, is that it is good to check. It is a relatively harmless tool to use to see if something weird comes up. While not the most powerful tool we have (which I would argue may be our own "common sense" to borrow an obnoxious but appropriate turn of speech) it is a tool. And when it is used properly, it can make the possibility of fakery compelling.

Perhaps it is time we requested specific guidelines on how to use it to prove fakery, even though this may discourage casual people from filling a thread with potentially useful information; the benefit could be that it would only do so when someone has actually found something. We could consider adding this to our required reading at the start of someone's registration.

Also, just in case you didn't really read the whole discussion that was moved there, you may have missed a funny little thing; 2013 and 2014 pictures of the Putin meeting, overlaid by Simon, results in this:

Image

Still think the event was as real as depicted in the media?
The photo above is clearly manipulated. EDIT: By that I mean the kneeling man was obviously added later.

While I didn't strongly contend that the photos of the recent Hollande / Merkel / Putin were real, I don't think it is safe to assume they are fake and this was my main point. If we had some evidence such as the photo overlay you show above, it would be certain of course.

My main point was to bring attention to the fact that it is difficult to use filters to detect fakery especially when JPEG compression can leave serious artifacts itself. I would like to figure out if it is mathematically possible to detect manipulation (I suspect it is in most cases) and then use mathematical methods for doing this. I am a bit of a programmer myself, and I have access to image editing tools as well as Mathematica which has a host of image analysis tools of a very advanced nature. My wife is also a mathematician and I bet I could get her to help me a bit.

I agree with you, that it can't hurt to check and that common sense is probably the best fakery detector. But I have to ask, because I am genuinely curious: what is the proper way to use the photo fakery detector tool? I checked the site when I brought this topic up earlier (last summer IIRC) and the exact algorithm is not detailed, nor is a satisfactorily deep explanation (to me but I am a bit of a skeptic) as to the exact methodology of its use.

This is why I think we need to have either a mathematical, programmatic, or at the bare minimum detailed logical basis for the detection of fakery. This might be a very deep undertaking but I am at least willing to brainstorm.

Thanks for reading what I have to say, and thanks again to all who post usefully here even if I disagree with some posts sometimes.
hoi.polloi
Member
Posts: 5060
Joined: Sun Nov 14, 2010 7:24 pm

Re: Forensic Photo Analysis

Unread post by hoi.polloi »

Sukiari, did you ever find out if there is a program similar to the PhotoForensics site, or something that could use Mathematica, which allows us to calculate digital manipulation?
Post Reply