Talk:Error level analysis

For Amateurs?

A search of Google Books shows a lot of professional looking books that discuss this technique. I understand one person may have called it for amateurs but that was an opinion, and quite possible biased as part of an attack on Higgins during a larger debate over the MH17 crash. I don't believe it's a reliable source for that particular statement. We have plenty of other sources that are more objective and don't seem to harbor the same view. -- GreenC 00:37, 28 July 2015 (UTC)

@Green Cardamom: To be accurate (which is probably a good idea here), the word 'Amateurs' is not used, while 'hobbyists' is. But it is a valid criticism that the article almost entirely relies on a single source. I have fixed that regarding the most serious limitation of the ELA, by citing a digital image processing conference paper from 2010 (i.e. from before the MH-17 allegations, so unlike more recent publications guaranteed to not be biased due to that tragedy). Please remove the NPOV template again. Thank you. PS. Interestingly, the cited paper proposes a improvement to ELA, a DCT-based PCA, but that seems out of scope for this article. Lklundin (talk) 07:05, 11 September 2015 (UTC)

@Green Cardamom: It has been more than two weeks since I notified you that I had started addressing your POV-concern regarding this article. You state that this article is related to Eliot Higgins, to which you have made subsequent, repeated contributions. Since you evidently have been contributing actively, but not returned to discuss this article I will therefore remind you that placing an NPOV-tag on an article carries with it an obligation to participate in the discussion here. So I invite you to reevaluate your view on this article, while I keep this quote from WP:NPOV dispute in mind: "In the absence of an ongoing discussion on the article's talk page, any editor may remove this tag at any time". Thanks, Lklundin (talk) 14:11, 28 September 2015 (UTC)

Sorry it slipped by. Regarding this sentence: "In May 2015 the citizen journalism team Bellingcat claimed Error Level Analysis as proof that Russia's Defense Ministry had manipulated satellite images from the Malaysia Airlines Flight 17 disaster." It's unsourced. Is it true that Bellingcat reached its conclusions on MLA17 based only on ELA? My understanding is they have many lines of evidence. -- GreenC 16:25, 28 September 2015 (UTC)

No problem - thanks for helping to provide the article. The amount of information being considered regarding the MH17 disaster is extensive, to say the least. It is my understanding that regarding this disaster Bellingcat has cited a range of sources of various kinds to reach their conclusions. However, specifically regarding their conclusion on the overhead imagery provided by the Russian Defense Ministry and ELA I rely on the phrase "The core of what they are doing is based on so-called Error Level Analysis" from the source cited here. Based on your comment I think the phrase could be moderated to say something like "as primary proof" instead of "as proof". You are welcome to suggest an alternative as well. If you also think it is necessary to explicitly attribute the statement, then that is fine by me. Lklundin (talk) 18:26, 28 September 2015 (UTC)

Then the problem becomes, is there are source that says this is the "primary" proof? Let me try something see what you think, I don't know what it will be yet, change if you want. -- GreenC 20:14, 28 September 2015 (UTC)

Ok I didn't change much mostly copyedit. Added "allegedly" since the people making the criticism of ELA are the same ones saying Bellingcat used ELA. Unless there is neutral evidence elsewhere in support.

The higher profile article (as evidenced by the 'discretionary sanctions' notice on) Eliot Higgins cites the same source and does not use "allegedly". So I think "allegedly" should only be introduced here if it is also introduced on the more important Eliot Higgins page. As for the "primary proof", I would say that we are getting into details, i.e. whether "primary proof" is an accurate representation of the sourced "core of what they are doing". Since you introduced ther NPOV-tag here, I think it is only fair that you act on any remaining POV-concern you might have regarding "proof". To conclude, I request that you drop the "allegedly" until it appears on Eliot Higgins and I suggest that you change (or propose a change) to moderate "proof" in accordance with the source cited here and on the Eliot Higgins page. Thanks. Lklundin (talk) 20:53, 28 September 2015 (UTC)

It's alleged by definition since it's the people making the criticism who are also making the claim. Eliot Higgins should contained alleged also. -- GreenC 01:48, 29 September 2015 (UTC)

Given your significant contributions to Bellingcat with its talk and your interest in ELA, I am a bit surprised that you were unaware that Bellingcat explicitly cites ELA as their method. As such I have largely reinstated my original formulation, now citing also Bellingcat. Lklundin (talk) 05:42, 29 September 2015 (UTC)

You're "surprised", huh? I don't track these things. I said above "Unless there is neutral evidence elsewhere in support," which you then provided. Thank you, I didn't know it existed. But now you've made this into an issue of bad faith, "I'm surprised". I have restored the wording -- without the word "alleged" -- it's clearer and more accurate. -- GreenC 13:15, 29 September 2015 (UTC)

What you think that I think about you has no bearing on the article. So please take a moment to reconsider your most recent article-edit, is it not consistent with the quoted sources. Lklundin (talk) 13:35, 29 September 2015 (UTC)

Obviously I don't see a problem so telling me to "read it again" is asking me to read your mind. Be explicit, I don't have time to guess what you think the problem is. -- GreenC 15:57, 29 September 2015 (UTC)

Before you made your most recent edit to the article, it cited sources to the effect that Bellingcat claimed to have revealed photo editing. This was consistent with the article on Eliot Higgins, which still has that information. The sources for this article are unchanged, but with your change it now says that Bellingcat merely were 'investigating' the photos, without actually concluding anything. When you ask to have this explained you appear at best as careless and at worst as disingenuous. So once again I ask you to reconsider your edit. Thank you. Lklundin (talk) 06:29, 30 September 2015 (UTC)

They used ELA as their core method to investigate the photos. They used multiple lines of evidence to make a conclusion about MH-17. We already went through this. Eliot Higgins is not a source and isn't a reason to do something in other articles - if anything that article needs to be changed. Look at this point your personal attacks are enough that I want to stop working with you. You clearly don't show good faith and attribute hidden motives ("disingenuous"). When people blatantly presume bad faith of me, I presume the same of them. I presume you have an anti-Higgins bias and are not a neutral editor. -- GreenC 13:41, 30 September 2015 (UTC)

You are confusing Bellingcat's general conclusions about MH-17 which are of no relevance here with Bellingcat's specific claim, sourced here, that ELA was used to reveal that Russia's Defense Ministry edited their MH-17 related satellite imagery. Lklundin (talk) 13:27, 6 October 2015 (UTC)

I looked at the Bellingcat source again and it doesn't say Bellingcat used ELA as a "core method" in determining the authenticity of the photos. It lists ELA as one of multiple lines of evidence. The only claims to "core method" are Bellingcat's critics. It we are going to assert something as fact, can't do it through the lens of a critic which is POV. To do so would require framing it like "According to XYZ, .." .. or use the word alleged with "core method" in quotes. Or drop core method entirely. -- GreenC 18:11, 7 October 2015 (UTC)

Correct. The Bellingcat source asserts that 'Error level analysis of the images also reveal the images have been edited', and this is exactly the information you removed and that edit is exactly what I have asked you to reconsider. So I would appreciate if you would reconsider your modification, which so clearly contradicts the source we are both referring to and which is still cited in the article. Thanks. Lklundin (talk) 14:58, 8 October 2015 (UTC)

The problem is our text is not restricted to only ELA it says "according to image forensics expert J. Kriese, Krawetz distanced himself from Bellingcat's conclusions" implying the Bellingcat's "conclusion" where based on ELA when in fact they were based on multiple lines of evidence. The way it was worded gives the wrong impression that Bellingcat made its conclusions about the photographs based on ELA when that is only partially true thus the factually correct and NPOV statement as it currently reads. -- GreenC 15:06, 8 October 2015 (UTC)

It is important to understand that in the context of this article, the conclusions that Bellingcat have reached regarding MH-17 in general are out of scope. What is relevant here is Bellingcat's use of ELA. So it is a step forward that you do not dispute the statement that Bellingcat claims that ELA revealed that the Russian Defense Ministry imagery was edited. With that in place, there is clearly no problem in citing Krawetz, who twittered: 'Yeah... chalk this up as a "how to not do image analysis"'. It makes no difference if Krawetz distances himself only from Bellingcat's use of the method he named ELA, or from all aspects of their image analysis. The point is: Bellingcat used ELA and Krawetz criticizes how they did it (and possibly everything else Bellingcat did in relation to image analysis). Lklundin (talk) 03:43, 10 October 2015 (UTC)

@Green Cardamom: I guess it is not fair that I expect you to do all the work, so I have rewritten the paragraph on Bellingcat, to make it less about them and more about ELA. I am quoting Bellingcat (only) about their own statement on their specific use of ELA. I am quoting the Spiegel article (only) about ELA. As something new, I am quoting the ELA author Krawetz that neither Bellingcat nor Kriese understand ELA. I invite you to give this attempt to improve the article a look. Lklundin (talk) 04:19, 17 October 2015 (UTC)

Clarifications

I'm tickled that someone wanted to write a wikipedia article on this. As a primary subject mentioned in this article (I'm Neal Krawetz), it is probably not appropriate for me to edit the article directly. So I'll just list a few items here.

1. ELA was first mentioned in my 2007 Black Hat Briefings presentation on digital photo forensics. Here's a link to the whitepaper and conference announcement: http://www.blackhat.com/html/bh-usa-07/bh-usa-07-speakers.html#Krawetz http://hackerfactor.com/papers/bh-usa-07-krawetz-wp.pdf

2. The link at the bottom should be to "fotoforensics.com" (plural) and not "fotoforensic.com" (singular). I registered the singular name domain because too many people (about 1%) forget the "s".

3. World Press Photo had many controversies that year. But I think the big one were the "independent experts". The "independent experts" may not have been independent. In particular, the chair of the WPP panel was the AP's Santiago Lyon. Hany Farid has given presentations with Lyon (e.g., http://video.mit.edu/watch/ethics-a-forensics-in-the-age-of-photoshop-photojournalism-7514/). Farid has blogged about how he does work for the AP (http://www.fourandsix.com/blog/2011/10/4/photo-forensics-from-eyes.html). Lyon wrote about how his company works with Farid (http://www.nieman.harvard.edu/reports/article/102763/Detecting-the-Truth-in-Photos.aspx). Lyon's company, the Associated Press, is a direct sponsor of WPP (http://www.worldpressphoto.org/sponsors). And the New York Times reported that Farid's company, FourAndSix provides software and services to the Associated Press and specifically to Lyon's division. (http://bits.blogs.nytimes.com/2012/09/18/taking-digital-image-forensics-out-of-the-lab-and-into-the-marketplace/).

In effect, we have a direct partner, contractor, and software vendor claiming to be an independent expert and offering support for their direct client. I point this out in one of my blog entries: http://www.hackerfactor.com/blog/index.php?/archives/555-A-Matter-of-Trust.html (If you're a direct partner/contractor/vendor, then you're not independent.)

4. Continuing the World Press Photo issues: I identified that the picture had undergone selective editing. The photographer was later quoted in an interview as saying that he did selective color enhancements in an attempt to generate a higher dynanic range. http://www.news.com.au/technology/photographer-says-his-2013-world-press-photo-of-the-year-is-not-a-fake/story-e6frfro0-1226642304141 QUOTE: "In the post-process toning and balancing of the uneven light in the alleyway, I developed the raw file with different density to use the natural light instead of dodging and burning. In effect to recreate what the eye sees and get a larger dynamic range. "To put it simply, it's the same file - developed over itself - the same thing you did with negatives when you scanned them."

In addition, a comparison against the initial published photo identifies selective editing: http://www.dn.se/nyheter/sverige/dns-fotograf-tog-emot-pris/ http://www.hackerfactor.com/blog/index.php?/archives/550-Angry-Mob.html

FourAndSix also made a statement that they never reviewed the publicly released image. They say that they only viewed the original, which was not available at the time the picture was awarded: http://www.extremetech.com/extreme/155617-how-the-2013-world-press-photo-of-the-year-was-faked-with-photoshop [QUOTE] Further confusing the situation, Fourandsix (which was co-founded by Hany Farid) told me on Twitter that the Raw image they analyzed matches the prize-winning photo — but he hasn’t compared the Raw photo with the original photo published by Dagens Nyheter, which is clearly different from the prize winner. Basically, this photo has been altered at some point in its life, but it still isn’t clear who altered it or when it was altered. [/QUOTE]

5. In response to this criticism, World Press Photo changed their rules. They now require unaltered originals for comparison prior to awarding their top prize. However, they also decided to not name their "independent" experts. http://www.hackerfactor.com/blog/index.php?/archives/659-All-For-The-Better.html

Using these new rules, WPP ended up disqualifying 20% of submissions due to excessive digital alterations. http://time.com/3706626/world-press-photo-processing-manipulation-disqualified/ The "Sports" category was so impacted, that no third prize was awarded. http://www.bjp-online.com/2015/02/image-manipulation-hits-world-press-photo/

6. Regarding lossless file formats, like PNG and BMP: A JPEG converted to PNG/BMP will retain the JPEG artifacts. ELA can be used to identify this type of file conversion.

Hackerfactor (talk) 05:57, 10 November 2015 (UTC)