Edited images: the proof

Questions, speculations & updates on the techniques and nature of media fakery
anon1911
Banned
Posts: 61
Joined: Mon Jul 25, 2011 2:57 pm

Edited images: the proof

Unread post by anon1911 »

Hello there!

At the Oslo thread I pointed out, that using more or less "unknown" algorythms on any image does seem to have some risks,
that I would like to define in this thread.

The reason why I opened this thread is, that I don't want to go too offtopic in the Oslo thread
and to just have a better place to talk about this topic from a more general point of view.

In my opinion this discussion fits right into this research forum.

Image

Let's start from this point:
So, you're saying that this color adjustment algorythm, which apparently estimates instructions from a given, but to you unknown, pool of data, actually prooves something?

The problems we have to face:
* We can only interpret the visual output of the algorythm
* By not knowing how the algorythm works on a small scale, we can't estimate it's output
* Some algorythms have the same name and do the same in general, but on a small scale the result is different from each other


Image

A (SIMPLICISTIC) LITTLE STORY

Imagine that some time ago you somehow broke the law and at that time the police took a nice photograph of your face to safe it in some sort of database in case they need it in the future.
Great, now imagine that the following picture is you, more specifically your face:

Image
Picture showing your face

One day you get a visit by two "friendly" policemen telling you,
that you have to come with them since there is evidence that you are involved in some sort of a crime.
Of course you didn't do anything and follow the policemen.
At the policestation they present you the following picture:

Image
Picture showing the person that commited the crime

Let's assume that this picture shows the face of a person, taken by a security cam, but it is too dark to be able to see who the person is.
You of course claim that it obviously is not you on that picture.
Now, what the police does is the following: They do some nice color adjustment and suddenly, it's you!

Image = Image ?

Image

HOW IS THIS POSSIBLE? (A MORE INDEPTH VIEW)

Okay, so let's take a look at the (very simplicistic) algorythm the police used.

Let's assume that the color adjustment that is used is something like a simple statemashine consisting of 3 states.

STATE 1: If input BLACK -> output is Grey
STATE 2: If input WHITE -> output is WHITE
STATE 3: If input GREY -> output is WHITE

Image -> Image !

So this is simply the way this speicific algorythm works.

The problem:
A certain image on which a certain algorythm will be used on will always give a certain result.
But the result hasn't anything to do with the original, there's no "higher" intelligence behind all that. The algorythm uses given instructions, which you generally don't know.
Just imagine they change one single state from the example above and the police would have never visited you because it's not showing you then.



Some questions coming to my mind:
Does changing images really proove anything?
How does this actually affect your/our research?


Image

EPILOG AND MY OPINION

Whenever I try to figure out wheter an image is fake or real, I always try to see if I could go to court with the information I have.
Just imagine, the guy from my story above goes to court and the question is: Does the picture on which the algorythm has been used on proove anything at all?
If I had to judge: No.


I'd be glad to see more people participating in this discussion.
burningame
Member
Posts: 109
Joined: Thu Dec 09, 2010 1:21 pm

Re: GETTING THE WORD OUT!

Unread post by burningame »

I confess to being a complete beginner in matters of image manipulation and video fakery, so please excuse my ignorance, and I hope I’m not barking up the wrong tree here.

Last few days I’ve just discovered the Hacker Factor blog, thanks to nonhocapito (“tools of the trade” thread). [http://www.hackerfactor.com/blog/] This guy at Hacker Factor, Neal Krawetz, echoes Simon when he says the best tools for analysis are a good pair of eyes and common sense:
…The hardest parts of forensic analysis isn't the tools; it's the training. Anyone can buy rubber gloves, swabs for collecting blood samples, and plastic evidence bags. But if you are not trained to properly collect, handle, and evaluate evidence, then the tools and methods are meaningless…

…While luminance gradient and error level analysis draw pretty pictures, the most important tool is basic observation. It is one thing to see the big and obvious signs of manipulation. It is something else to remember all of the fine details.
There seems to be quite a few tools available to analyse images: gradient map test, Error Level Analysis, Principal Component Analysis, Blur Detection, Color Distance, etc. These tests throw up ‘pretty pictures’, as Krawetz says, but they also highlight the fact that there are all kinds of hidden clues embedded in the photos. And that’s the clincher – the stuff you can’t see, being revealed.

To me one of the most effective of Simon’s image deconstructions - which proves to even the slowest observer that the photo is fake - is the one of the firefighters where Simon applied the gradient map test:

http://www.septclues.com/SIMCITY/RUBBLE ... pTest1.gif

This kind of demonstration is very effective indeed.

So I’m wondering, (1) how come there aren’t any other analyses of images as per Krawetz on this site, with the exception of the fireman gradient map (unless I’ve missed a whole section here); and (2) can these tests or something like them be applied to video as well? I mean, is it as simple as taking a frame out of the video and performing these tests, or would that be meaningless? (I have a feeling that’s a supremely dumb question, but that’s where I’m at at the moment).

However, with a view to trying to convince the skeptical AND image-blind public (and here I am talking just as much about myself, or maybe as I was one year ago), I thought that an analysis with these different tools might make it easier for the layman to realise what is actually going on. Do you think we need a Vicsim-style report, but maybe shorter, focussing on two or three of the most obvious photoshopped examples – the cartoon plane, for example and other iconic images. I know Simon has totally debunked the images ‘from the top’ i.e. with naked eye, but wouldn’t a successive analysis, on many different levels with all these tests, where you can ‘see’ the fakery, be worthwhile? I just think it would add an extra level of scientific thoroughness.

I would like to have a go and try fooling around with some of the classic images, just for my own education, seeing what happens when you run these various tools. Can one download an image that’s as close to an ‘original’ as possible? Or is that the problem – there are no ‘originals’, the images have been so downgraded that they will simply not respond to this kind of analysis?
nonhocapito
Member
Posts: 2579
Joined: Sat Jul 10, 2010 5:38 am
Location: Italy
Contact:

Re: GETTING THE WORD OUT!

Unread post by nonhocapito »

Burningame wrote:To me one of the most effective of Simon’s image deconstructions - which proves to even the slowest observer that the photo is fake - is the one of the firefighters where Simon applied the gradient map test:

Image

This kind of demonstration is very effective indeed.

So I’m wondering, (1) how come there aren’t any other analyses of images as per Krawetz on this site, with the exception of the fireman gradient map (unless I’ve missed a whole section here); and (2) can these tests or something like them be applied to video as well? I mean, is it as simple as taking a frame out of the video and performing these tests, or would that be meaningless? (I have a feeling that’s a supremely dumb question, but that’s where I’m at at the moment).
[Burningame, I moved this post of yours into this thread started by anon1911, and the thread itself into the "question of fakery" forum.]

Only recently i was proposing I think on the Oslo thread that we should have this discussion... I am not a particular expert on photo manipulation myself, so like you I too am in search of standards, to clarify to myself and others what and how exactly digital manipulation of pictures can be demonstrated. (I'd leave out videos for the time being, only to simplify the discourse, although it is obvious that most of the same investigative criteria, plus specific others, would apply.)

Of course to have this discussion we must all agree on the fact that there is not one method nor one way to get to proof without doubt, but more often than not there is a visual hunch supported by a number of technical clues, clues that sometimes taken one by one would are not enough.

The point, as always, is in cumulative evidence, which I believe is the same way "scientific experts" work with forensic analysis of images: cumulative clues.

I know the whole thing may sound as a too challenging, hopeless task, but if done right, a thread like this one could turn out to be very useful, outside the boundaries of this forum as well. I am hoping for other expert members to post their ideas on this...
Maat
Member
Posts: 1425
Joined: Thu Sep 09, 2010 1:14 am
Contact:

Re: Edited images: Any proof at all?

Unread post by Maat »

Anon1911,

I'm sorry, but I find your analogy neither appropriate nor accurate in this context and could even be inferred as an attempt to confuse and mislead those new to the subject. It also doesn't help to convince anyone that you understand the processes you dispute by persistently misspelling the apparent keyword of you argument, algorithm, as "algorythm" (10 times in this post). I don't see how it could be excused by German being your "mother" language either, since the German word is Algorithmus, also with no Y.

As you may recall, I asked you the following question on the Oslo thread (here) where you began your "algorythm" argument to dispute Makkonen's photo analyses. So I'll ask it again;

Is it really necessary to know every technical detail of how a program works in order to observe results that are consistent with known parameters and proven by comparison?
If you want to know about jpg vs other file types, for instance, you can read the expert info demonstrated on this site:
http://www.hackerfactor.com/blog/index. ... toria.html

Did you read the detailed information and examples clearly demonstrated on that site?

My point is that one does not have to be a photographer or graphics expert to see when an image alleged to be real does not match one's reference frame of reality, i.e. it looks wrong.

Therefore, digital analysis programs, when understood through practice and experience by those who use them & comparative examples demonstrated, can only help to confirm, i.e. prove, what we could already see subjectively.

I agree with nonhocapito that this is a forensic process that involves following diverse clues that lead to logical conclusions from cumulative evidence, just as the professionals do.

Such a preponderance of evidence, including acknowledged experts' opinions, is acceptable in courts of law. ;)
Heiwa
Banned
Posts: 1062
Joined: Tue Oct 20, 2009 6:20 pm
Contact:

Re: Edited images: Any proof at all?

Unread post by Heiwa »

Image
That above picture is edited is quite easy to see. The picture, snapped just when a WTC tower is destroyed and before the photographer is blinded by the dust, consists of three parts.
In the foreground is an old building at the tip of Manhattan and some trees. Behind the building is a lot of smoke and dust because one of the WTC towers has just been destroyed and is becoming rubble. And in the upper right corner we see a glimpse of a big undamaged top, outer wall section just below the roof of the WTC tower that is still falling down from 400 meters height … with smoke (??) still escaping from its undamaged front (so there must be undamaged floors, offices behind the wall). However, it is quite clear that the wall section is just copy/paste from the movie of the tower being intact/burning with smoke excaping from the wall, etc. So the picture is edited. You wonder why. To show that an intact top part crushed the stronger bottom part?
nonhocapito
Member
Posts: 2579
Joined: Sat Jul 10, 2010 5:38 am
Location: Italy
Contact:

Re: Edited images: Any proof at all?

Unread post by nonhocapito »

Dear Heiwa I see you wanting to plug your perfectly reasonable demonstration of the fallacy of the towers collapse, but, as you obviously understand, we are not discussing what in general makes a picture fake and unreasonable and obviously manipulated.

Clearly the larger part of the analyses conducted on this forum infer the fallacy of pictures by simply looking at what they represent, and the absurdity of the story they try to tell. Like in the case you just presented (although there probably also are digital elements at play that could be discovered in it: starting from noticing how the whole left side of the picture was stretched by copy-pasting, or how the exif date is "2002:09:11"...).

We are trying instead to discuss the digital clues that can tell a manipulated picture from a regular one, regardless of what the picture represent.
Traces of layering composition, manipulation, rendering of elements, in other words an effort to possibly standardize a process of "reverse engineering" on pictures, to understand how they are put together in the first place.
Heiwa
Banned
Posts: 1062
Joined: Tue Oct 20, 2009 6:20 pm
Contact:

Re: Edited images: Any proof at all?

Unread post by Heiwa »

nonhocapito wrote:Dear Heiwa ... We are trying instead to discuss the digital clues that can tell a manipulated picture from a regular one, regardless of what the picture represent.
Traces of layering composition, manipulation, rendering of elements, in other words an effort to possibly standardize a process of "reverse engineering" on pictures, to understand how they are put together in the first place.
A picture always represents something or other and my example above is an obvious fake ... but not standardized. Just an idiot trying to be clever. We don't even know who she is! Must be a she because it is so stupid.
Here is another example:
Image
The original is by Richard Drew. I have added some comments on it. Richard and wife stills live at NY and Richard has worked for AP 40+ years. I wonder how Richard snapped this one. Or photoshopped it. Or edited it. Any clues ... except pure corruption?
burningame
Member
Posts: 109
Joined: Thu Dec 09, 2010 1:21 pm

Re: Edited images: Any proof at all?

Unread post by burningame »

I think I've just realised something - can anyone please confirm or elucidate: that there are no PHOTOS of any of the 911 iconic images - just frames extracted from the 911 videos. That is, no one 'took a photo' of any plane - everything is VIDEO. So even if you got the 'cleanest' copy of the videos, if you extract a frame, there's not going to be any kind of embedded evidence of manipulation that would show up in these various tests – ELA, PCA, gradient map, etc.

That’s why Simon has only done this kind of testing with a photo (as above). We only HAVE the naked eye, and knowledge of what’s possible.

Am I right?
nonhocapito
Member
Posts: 2579
Joined: Sat Jul 10, 2010 5:38 am
Location: Italy
Contact:

Re: Edited images: Any proof at all?

Unread post by nonhocapito »

Burningame wrote:I think I've just realised something - can anyone please confirm or elucidate: that there are no PHOTOS of any of the 911 iconic images - just frames extracted from the 911 videos. That is, no one 'took a photo' of any plane - everything is VIDEO. So even if you got the 'cleanest' copy of the videos, if you extract a frame, there's not going to be any kind of embedded evidence of manipulation that would show up in these various tests – ELA, PCA, gradient map, etc.

That’s why Simon has only done this kind of testing with a photo (as above). We only HAVE the naked eye, and knowledge of what’s possible.

Am I right?
Well the balance between video and pictures seems to be in favor of video, but there are plenty of "photos" depicting the impact or the collapse. But I think, since most of the material is being re-rendered and re-mastered, now they have a chance to make sure the manipulation shows as little as possible. It's not the time of the devil faces anymore.

This is a "photo":
Image
From http://www.corriere.it/reportage/esteri ... 8.shtml#10

It is credited to "AP Photo/Chao Soi Cheong" , who is also responsible for a shot taken a second before:

Image
From http://cryptome.org/info/ap-911/ap-911-pack.htm

Not very believable... anyway "Chao Soi Cheong" must have made so much money with these two shots he retired, cause he doesn't seem to be responsible for anything else but this 9/11 crap: http://images.google.com/search?q=Chao+ ... 78&bih=914 :D

just for the f of it, here is a gif depicting the two shots in sequence:
Image
simonshack
Administrator
Posts: 7339
Joined: Sun Oct 18, 2009 8:09 pm
Location: italy
Contact:

Re: Edited images: Any proof at all?

Unread post by simonshack »

nonhocapito wrote:just for the f of it, here is a gif depicting the two shots in sequence:
And here's the video-version of it. Uh, well...not quite the same angle. But I hope the absurdity of it all is obvious to any keen observer:

Image

And did "Chao Soi Cheong" (or his videocam-equipped colleague) just hang around in that location for 56 minutes - until the WTC2 collapsed?
Image

^^^^^^^^^^^^

I will, in due time, put together a comprehensive response to Burningame's questions - here on this thread.
pov603
Member
Posts: 870
Joined: Thu Jun 30, 2011 8:02 pm

Re: Edited images: Any proof at all?

Unread post by pov603 »

Just a quickie.
http://uk.news.yahoo.com/first-photo-ip ... 00246.html
Above is a link to a yahoo report on a photo taken [allegedly] from an iPhone 5.
My reason for posting this is to highlight the fact that Yahoo! are pointing out that the report states: 'We love this story. Would any other phone, other than the iPhone 5, have enough pull to make people go to the effort of sorting through EXIF photo data, or partaking in heavy image processing - just to see barely visible reflection of what might turn out to be the next-gen Apple handset?'
Now i have not known* anything about exif/EXIF [capitalise or not?] until coming on this forum but have gathered that there is some useful information to gleam from it whether it is there or not there as the case may be.
i just wonder is someone somewhere at Apple or Yahoo! trying to get some message across regarding information posted with photos?

*Edit: Just correcting my spelling... :rolleyes:
Last edited by pov603 on Sat Sep 10, 2011 11:33 am, edited 1 time in total.
Maat
Member
Posts: 1425
Joined: Thu Sep 09, 2010 1:14 am
Contact:

Re: Edited images: Any proof at all?

Unread post by Maat »

Ah, now that is interesting, pov.
Not only for the hint about what is/isn't reliable about Exif data, but as another advertizing teaser for Apple's iPhone 5 (that's been 'in the works' for ages now!) and all over a photo of some sushi. Very fishy, indeed! :lol:

However, what I find really interesting about the cellphone camera 'developments' in general, is how incongruously poor the comparative photographic quality and results are, despite all the superior electronic refinements & computer wizardry in everything else <_< Seems very deliberate to me — being so convenient for disguising image fakery in PsyOps skullduggery as "amateur" video from a "bystander's" camera phone. :rolleyes:

Since they obviously wouldn't want to give up such a handy asset in their PsyOp 'toolbox', any new 'improved' iPhone 5 (if released in the next month as supposedly 'planned') must be in sync with whatever the perps have 'in the works' too. :ph34r:
Jonathan
Member
Posts: 100
Joined: Tue Jul 26, 2011 8:17 am

Re: Edited images: Any proof at all?

Unread post by Jonathan »

FWIW: EXIF is Metadata written by the camera to annotate the image...
It might be correct but it is just as easy to change it, or to add/remove part or all of it as it is to write these sentences.
Altering it is no magic.
Maybe a clue, but no proof for anything at all.

here is one program
and here another
just the first hits in a search for "edit exif"
burningame
Member
Posts: 109
Joined: Thu Dec 09, 2010 1:21 pm

Re: Edited images: Any proof at all?

Unread post by burningame »

I’m not talking about Exif data. What I’m getting at is this: according to the Hacker Factor blog, [http://www.hackerfactor.com/blog/], there are various tests that can be performed on photos, that is:

Error Level Analysis
Principal Component Analysis
Color Distance Algorithm
Blur Detection
Luminance Gradient
Perceptual Hash Algorithm
Gradient Map Test

They detect tell-tale clues that a photo has been manipulated, that unlike exif data CANNOT be erased or altered. (I refrained from cutting and pasting the many examples from Krawetz’s blog to illustrate; I figure we can all go to a URL – but I will if you need it.)

My question is: why can’t we use these tests on all the iconic images (the ‘best’ ones which would be familiar to the public) to prove beyond any doubt that they were manipulated? Wouldn’t the added scientific proof only help the case for media fakery?

[EDITED: removed 'Color Distance Algorithm' - listed twice]
Last edited by burningame on Thu Sep 15, 2011 3:19 pm, edited 2 times in total.
anon1911
Banned
Posts: 61
Joined: Mon Jul 25, 2011 2:57 pm

Re: Edited images: Any proof at all?

Unread post by anon1911 »

I'm sorry that I didnt answer to all your messages. I will when I find the time to.

This statement interests me though:
Burningame wrote: They detect tell-tale clues that a photo has been manipulated, that unlike exif data CANNOT be erased or altered. (I refrained from cutting and pasting the many examples from Krawetz’s blog to illustrate; I figure we can all go to a URL – but I will if you need it.)
Considering the fact that each image, no matter what the file format may be, is just a bitstream, please tell me how you can be sure of the fact that this stream cannot be rearrenged? Rearrenged in a way to make the manipulation undetectable?
(Of course I'm talking about clues that are only visible by altering the image with a certain 'analysis')

I'm pretty sure that this is actually possible.
Post Reply