Tuesday, July 05, 2011

Obama's long form birth certificate may have been a forgery

Update on Wednesday: democracy in America may still be operational. A Hawaii court has ordered a new investigation into the certificategate and asks the Hawaii Department of Health to produce the original copy of the birth certificate: a subpoena.

Update: The best alternative explanation to forgery I have heard of is that the file used to be a DjVU file at some point. See a description of DjVU compression that could explain many things. I reproduced most the anomalies via DjVU encoders: see a new article about Obama and DjVu. Meanwhile, here is a high-res JPG picture of the certificate (without the green junk):

Shift-click to zoom in.

I used to be a lukewarm birther but when Obama's folks released the
long form birth certificate (key PDF for this article),

White House April 2011 press conference (announcing the PDF file)
I lost most of my interest. The birthers gradually began to resemble those who doubted moonlanding. Well, the analogy surely wasn't that perfect but I did decide to wait for another good reason to doubt that Obama's Hawaii birth certificates are legitimate.

Click to zoom in: a hugely magnified long form birth certificate. Aren't some letters missing? Did they ever fit together with the letters above? That's the most ridiculous thing I've ever heard. Details will be discussed below.

Well, the surprising event came today. I received a Tea Party Telegram by e-mail that claimed to have new evidence that the certificate is forger. I thought: well, it's some kind of conspiracy rubbish again. However, they claimed to have an analysis by Mara Zebest, a photoshop expert with quite some experience:
YouTube: Zebest's 3-minute introduction
YouTube: Zebest's 43-minute discussion
Zebest's written final draft

If you want to investigate this issue seriously, I ask you to download a PDF reader. For the sake of compatibility with your humble correspondent, also download Google Chrome at chrome.google.com. It has a built-in PDF reader that I found particularly convenient.

Inconsistencies in pixelation, anti-aliasing, colors, and layers

Zebest's written analysis - that is linked above - describes many problems but it's just fun if you can rediscover all those things yourself. So please, open this blog entry in Google Chrome and open the long form birth certificate from a White House server in your Chrome built-in PDF viewer. If it opens in Adobe or NitroPDF or anything else you may have, go to about:plugins (this link doesn't work, sorry), distable all those PDF readers, and activate the Chrome built-in PDF viewer.

I actually think that I have found some even more spectacular problems with the PDF file than Ms Zebest so please be patient. Well, the Czechs have had quite some historical experience with forgeries of historical manuscripts produced to support naive patriotism so my American readers shouldn't be surprised that I approach this situation from the position of the experienced one.

Now, let me remind you about the basic rules of the game. This is supposed to be a scanned real document printed on paper in 1961. The greenish pattern was added by the White House to protect their privacy - I have no idea why it helps - but this pattern isn't too important and has nothing to do with the issues that will matter.

Focus your attention e.g. on the line 6c, the hospital. Click at the "zoom in" (plus) button in the right lower corner of the Chrome PDF viewer about 10 times. You will see the following:

Click to zoom in

If you have a regular monitor - laptop is pretty bad - you will see that pretty much the whole text "Name of Hospital or Institution" is written in nicely monochromatic colors - dark green on white with sharp transition between the dark and light pixels.

However, some letters are neither monochromatic, nor sharp, nor dark green. They're gray, continuous, and anti-aliased (and without green, as you can see on a good enough monitor). Well, I have recommended you to use the Chrome PDF viewer because something remarkable happens when you press the "zoom in" (plus) button eleven or twelve times. Here is an example what happens:

Click to zoom in

Cute! Some of the letters have nicely disappeared. Clearly, those letters - which were anti-aliased and not sharply monochromatic - are in a different layer and Chrome helpfully eliminates this layer altogether. The silly green background has disappeared as well but it's not the issue here. The individual missing letters, e.g. "N, H, I" in "Name of Hospital or Institution" are much more interesting.

An obvious explanation is that these letters were artificially and digitally added and modified to the image - probably because someone didn't like how these letters looked in his previous version of the certificate. I don't claim that I possess a rigorous proof that there isn't any alternative explanation but I am surely curious about your alternative explanations why some of the lettes are missing.

In fact, the detailed review of the letters that are missing in this layer paints a very tantalizing, self-consistent story. Why don't we look at all the missing objects in this layer? ;-) The following articles will resemble a "spot 5 differences" game.

Click to zoom in.

The first line contains the identification number of the certificate. It should be 151-61-10641. Helpfully enough, Chrome PDF viewer, 11 times magnified, completely erases the last digit "1". The birthers have pointed out that the identification number for Obama, who was born on August 4th, 1961, was strangely higher (by 3 or 4) than the identification number of some Nordyke twin sisters who were born one day later.

Shift-click to zoom in. Nordyke twin sisters' birth certificates. They were born in the same hospital just a day after Obama - maybe - but their numerical ID is actually smaller by 3 or 4, respectively. Despite the lower ID numbers, their certificates were signed by an attendant 3 days after Obama's, on August 11th. There are many differences between their and Obama's certificate.

Chrome PDF viewer shows you that there is something wrong about the last digit. It doesn't fit there. Because the last digit "1" belongs to a different layer of the digital PDF document, it couldn't have been scanned with the other sister digits on any "parent" document on paper. Instead, whoever has photoshopped the long form birth certificate has added the final "1" now, in the 21st century, by digital methods.

Again, if you think that you have a different explanation why the final figure "1" appears in a different layer of the PDF document, I am curious about your alternative explanation. By the way, it's plausible that I am the first person in the world who noticed that individual letters and digits belong to different layers. But I realize that the multiplicity of layers was noticed by Ms Zebest.

By the way, the people who don't use the helpful feature of the Chrome PDF viewer may still notice that the final "1" in the identification number has a significantly differently color, level of smoothness, and antialiasing properties than the other digits. But its belonging into a completely different layer seems to be a much simpler way of showing that it was added.

Remarkably enough, the Chrome PDF viewer, when fully magnified, shows you the version of the forged document at an intermediate stage of the photoshopping which is very helpful.

Let's continue to look at the missing letters and digits in this layer. In the next line, "BARACK" only appears as "BA ACK" - the letter "R" was added at a later moment. I won't include the screenshots for all these details because it's obvious how they look like. The letter "R" is antialiased while the rest of the first name is sharp and monochromatically dark green - so you don't need Chrome PDF viewer to see that something is wrong with the letter "R". In the preliminary version, the letter probably looked badly so they pasted a letter "R" from a different font over it.

Also, the digit "1" identifying the first field and the period after the corresponding "2" are missing in this layer. The letter "g" is missing in the word "single" in item "3". Most likely, their original font used a version of "g" that looks like "9" but they needed the version of "g" with two disks connected with a tilde in between.

Other missing things are "p" in "triplet", comma after "4" (birthday), and the strange different-font "m." in P.M. below "hour". "6a" is missing in this layer. I forgot to say that all the separating lines are missing, too. I have already mentioned the missing capital letters - and a few others - in "Name of Hospital or Institution".

A funny twist occurs on the next line. In "Usual residence of mother", "d" is missing in "residence". There's some extra story here. Ms Zebest has noticed that some of the letters "d" in the would-be original printed form are "serifed" at the bottom right while others are not - compare the words "Inside" and "judicial" in entry 7e. So the author of the PDF file probably noticed that this single "d" was serifed while others were not, so he or she manually eliminated the serification at the bottom - but didn't do so everywhere consistently.

I could continue with lots of letters and digits that are missing in the layer and that are different - "K" in "KENYA" as well as "Kansas", "S" in "STANLEY", and so on, but I am sure you can see it yourself.

So far, you could have thought that the sharply pixelated, black-and-white layer was obtained by scanning a real document, without further manipulations. However, here comes another observation by Ms Zebest. It's my favorite one. Look at these letters in 7e:

Click to zoom in.

Do you see something unlikely here - for a scanned document? Look at the last "i" in the word "Inside" and the first "i" in the word "judicial". They're identical: pixel by pixel! Try to estimate the probability that this occurs by chance. Compare any pair of letters "i" - and you will see roughly dozens of pixels in which they differ. There are dozens of pixels in which they may differ - because of noise affecting the old paper, the strength of the printing or typesetting, as well as imperfections of the scanning.

I think it is more or less implausible that these two letters "i" are pixel-identical to one another by chance. They had to arise from digital copying of a common ancestor. If you have an alternative theory, I am once again eager to hear it.

And it's not just one pair of perfect clones. Look e.g. at the word "judicial" in 7e. Right before the letter "i" we just discussed, there is the letter "d". It is pixel-identical e.g. to the letter "g" in "Residence" in 7g, including the funny and illogical hole in the middle of the vertical line (on the right side of the letter). I guess you will be able to find many and many pairs of exact clones. ;-)

Just to be sure, this funny "d" with the hole on the right vertical actually has very many pixel-identical clones. For example, in 6d, all four copies of this "d" are pixel-identical to it - see the letters in "6d" itself, "Inside", "judicial", and "district". The probability of such an agreement to occur by chance is something like one part per trillion. Another identical sibling is the first "d" in "Address" in the field 7d and two more are the two d's in "Address" in 7f. And look at the "d" in "Residence" in 7g. A trillion was just raised to a quadrillion. ;-) Also "kind" in 12b and "d" in "Industry" in 12b. And "maiden" in 13. And "d" in "Outside" in 17a. And "d" in "[A]ttendant" at the bottom part. "Alive on the date" and "midwife", too. Quintillion and more.

In the line 20, the date stamp "AUG-8 1961" has the final "1" typeset in a clearly different, darker color than the much lighter previous digits of the date. If you suggested that it just happens that all the dozens of pixels in the final "1" are exactly black while all the 100's of pixels in the other digits are exactly gray, it's insane but OK. Someone could have carefully used a pencil to make the final "1" darker, couldn't he?

However, I may also show you the opposite kind of an anomaly or mistake. In the line 22, the creator of this PDF file needed to make "AUG-8 1961" lighter (gray) once again. So he has clicked to change the darkness of these characters. However, "1", "9", and "1" in "1961" stayed black. Even more seriously, when he clicked, the text "Date A" in "Date Accepted" above became light gray, too - as if it were printed by a stamp, too. :-) A truly cheap mistake.

Well, it's mistake that would be hard to avoid, too. The forger probably needed to make the stamps lighter so he used a "bucket tool" to fill in the color - but because the "Date A" letters were continuously connected with the letters in the stamp, they got colored gray, too. Again, I am curious about your alternative explanation of this point. After 100 comments in the comment thread, there's no alternative explanation.

Click to zoom in. The date should have been made light gray because it's a stamp. However, the careless creator of this image was probably color-blind so he kept the digits 1,9,1 in 1961 black - and on the contrary, he made nearby "Date A" (which shouldn't be a stamp) lighter even though it should have stayed black. ;-)

So even this black-and-white, sharply pixelated layer with missing layer was digitally engineered, not scanned. It was further manipulated in the other layers. If you look at the naked black-and-white layer, it's clear that it doesn't resemble a real document (the sharp pixels and black-gray-white 3-color setup is extremely unnatural for a scane), so the fake greenish background pattern was added to make it superficially look like a real scanned document and to mask the narrow-minded selection of the colors, not because of the bogus "privacy" issues.

Needless to say, Obama couldn't have hired the whole State Department or Department of Homeland Security to prepare this forgery. He couldn't have said - Hey, Hillary, could you please ask all your best people to work on my long form birth certificate? The certificate had to be created by a very small circle of the faithful only which is why it is such an amateurish piece of work.

Again, if you have an alternative explanation of the anomalies, I am curious about this explanation!

Bonus: huge identical letters

Here is my so far most robust proof that the image inside the PDF file can't be a scanned document.

Click to zoom in and carefully look at the image in a new window.

The picture above that you should magnify isolates the title of the document - "certificate of live birth" - at the top of the White House PDF file.

To start, compare the two letters "I" in the word "certificate". You see that both of them have many teeth - these are really imperfections - on both of their vertical boundaries. The height of the letters is roughly 40 pixels - and the letters "I" have two boundaries. So each of the 80 "disputable" pixels or so can be either black or white. The pattern is somewhat analogous to keys that we use to unlock the apartments - and there are 2^{80} different inequivalent keys.

Indeed, the two random keys - two letters "I" in the word "certificate" - are completely different and kind of random.

But now, compare the first letter "I" in "certificate" with the letter "I" in "live". They're pixel-identical! The probability of such an agreement by chance - if they were obtained by scanning two different "i"'s on a paper - is 2^{-80} which is about 10^{-24}. But they clearly do agree. You don't need to rely on any birther. Take the PDF file from the White House and magnify it 11 times in Google Chrome.

If you wanted to suggest that it is a coincidence that the teeth on these two letters "I" exactly agree, I can shock you again. I can shock you as many times as you want. Look at the first "E" in "certificate" and on the "e" in "live": they're pixel identical once again.

Derek's PDF

The PDF file says - in a text editor - that it's been created by a Derek. Derek R.B. Douglas from the White House Domestic Policy Council is a top candidate to be this Derek - he could have been at least the last person who manipulated with the file.

The nine images inside the PDF file

If you order e.g. NitroPDF viewer to "extract text", you get an empty file. If you ask the viewer to extract images, you get 1 JPG file and eight transparent GIF files. The JPG file is the only colorful one; the GIF files are all black-and-white with different interpretations of the foreground color (different levels of gray). After 100 comments, no one can even explain why a scanning/conversion/compression program - and which one - would separate the image into a colorful JPG and transparent black-and-white (1-bit) GIF bitmaps. ;-)

The JPG file looks like this:

Click to zoom in.

Note that this JPG file contains the artificially added green background (which is not curved in the same way as the text, check that it's horizontal under "State of Hawaii" in the upper left corner even though it should be slanted) *together* with the letters that are missing in the black-and-white layer I have discussed, some signatures, and a few other things. Despite the huge number of layers, the only admittedly artificial part of the image - the green pattern in the background - apparently didn't deserve its own layer!

Here is the first and main GIF file:

Click to zoom in.

It's saved as PNG on my Picasaweb but it is originally lossless GIF when extracted from the PDF. Don't overlook that this main GIF image contains not only printed letters but also "graphics", at least the parents' signatures. Well, a part of them. Handwritten "Ann D" appears on the green JPG while "unham" appears here - all these separations are strange and don't look too "automatic". Both signatures, "unham" and "Obama", look too perfect: the upper and lower boundary of every single "handwritten" letter is on the same horizontal line, within a 1-pixel error.

If you wonder, the sizes of the pixels in different layers differ, too. That means, among many other things, that "D" has a lower resolution than "unham", suggesting that the latter was added at a different moment. Alternatively, a scanned picture had to be evaluated by a software that randomly chooses different resolutions for different parts of the image.

The second GIF file extracted via NitroPDF is the signature of the registrar:

Shift-click to zoom in into another window.

This is a lossless GIF image and you may see that the letters in it are often identical. For example, compare "I" in "IS" in the first line with the two letters "I" in "Hawaii" on the third line. They're pixel-by-pixel identical. A special compression algorithm that looks for clones would be needed to create this out of a legitimate picture.

Another GIF image is the April 2011 date.

The next GIF image is a 1961 date stamp with 1,9,1 missing in 1961.

One more stamp with a missing digit...

"Non" deserves a special GIF image, too. :-) If you wonder where this "Non" appears, it's a part of "None" - the answer to the question what's Ann Dunham's occupation outside home during pregnancy. The remaining "e" appears on the black-and-white main page. Those who claim that this PDF file was created automatically must also explain why "Non" got a special layer - and was included neither in the greeny background JPG file nor in the main black-and-white GIF file with the readable letters.

If there were an original long form document that was digitally edited, does this separate layer suggest that there was something strange about Ann Dunhan's occupation?

A couple of nearly invisible spots... Well, they're visible on TRF because of our dark background but this light gray color appears on a white or transparent (and mostly light beneath it) background in the PDF form.

The same caption - spots. The absence of any larger or non-spherical mottles on the two pictures above could be viewed as unnatural, too. I would even say that it looks like that these dots of dirt were added by the "pen" tool of a Photoshop by a few touches with a fixed value of the radius. ;-) It's also strange why this dirt is concentrated in two small rectangles of the document only.

And an even more elementary question is whether an automatic scanning/compression program could be able to identify oil/dirt on the document and put them exactly into a separate layer, instead of including them to the colorful JPG picture. It's quite some reverse engineering! Again, I don't have a rigorous proof but "intelligent design" looks like a more plausible explanation. The oily spots were just an extra, isolated phase of a project attempting to create a document that looked realistic to its author.

If this separation to a JPG and many GIF files with all the properties were done automatically by a program and you know which one, I would like to know which one and, if you claim to know the answer, I would like to you to send me a similarly scanned and processed document that will produce a similar collection of images if I extract them by NitroPDF. There has to be a JPG image and GIF files and the GIF files must contain anomalously pixel-wise identical segments for many letters. And many other strange things discussed above.


  1. Would explain why he took so long to release it. The possibility that Obama is a huckster is truly depressing.

  2. In Australia we have a Prime Minister born in Wales. Anyone can aspire to be the country's leader.
    In the USA they have a rule that you have to be born in the USA.
    In Australia we have a Prime Minister who campaigned that there would be no carbon dioxide tax and when installed in office changed her mind. Her position is based on fraud.
    It appears that POTUS's position is also based on fraud.

  3. None of these "anomalies" are indicative of a forgery. In fact, the anomalies and the reason why they exist in the PDF were noticed and explained by Dr. Neal Krawetz, who has extensive experience with computer forensics and image analysis, as soon as one day after Obama released his long form certificate.

    See http://www.hackerfactor.com/blog/index.php?/archives/428-After-Birth.html

  4. Regarding my previous comment, since it was done several months ago and does not directly address Zebest's analysis, you may be interested in this analysis, which is specifically about Zebest's claims.


  5. There's an intermediate shade around some of the letters, but still they're so harsh and jagged around their edges that they look as if the image containing them was faxed at some intermediate or early point in the processing.

    Back when Rather had the fake memos I noticed the (alphabetic) letters' ungradated and jagged edges and it reminded me of a fax that I had had occasion to examine minutely some months earlier. But at the time I didn't call it definitely a fax, since I assumed that a cheap photocopier might do the same. Later on it turned out that the fake memos were indeed faxed. I didn't realize that it would need to have been a very cheap photocopier indeed. A fax imager is monochromatic because monochrome is the standard mode for faxes.

    It's been years and I'm not sure that the long-form certificate images' letters have the same "style" of jaggedness that I saw in a fax and in the fake memos. But I can't help wondering if the certificate images too were faxed. (I would strongly assume that Bill Burkett did not fax them! :) )

  6. Dear antpogo, I don't see things in your analysis that would explain my concerns - and indeed, most of them were already mentioned by Zebest.

    Your text starts with an ad hominem attack on Zebest. No doubt, she is not an expert on anything related to computers. She was surprised to hear that one may open a PDF file in a text editor to see some extra information not seen visually.

    But she still saw some points. You don't really explain where the different color structure of the different layers comes from and how could they have been divided into these layers automatically. And you don't explain the bit-identical segments in various pictures.

    Some of your comments are literally upside down. For example, you use the existence of the two final GIF images with spots as an argument that it was not forged. WTF? Could you please explain me a logical reason why an algorithm that divides the scanned picture into layers automatically wouldn't include the two GIF files with oily spots into the JPG background?

    I want to see a PDF file of another, real, scanned document that has the same strange qualitative properties as this one and that was prepared by an automatic manipulation that wasn't treating individual portions of the documents separately.

  7. Dear Geoff, well, Australia is still a country with a formal head who is the British queen so it would indeed be strange it you didn't allow women from Wales who moved to Australia to become top politicians.

    Still, this feeling is not mutual - an Australia-born person won't become a queen or king of Britain haha.

    Otherwise I agree she was not telling the truth.

  8. A brief addition to my previous comment, since I just noticed the update regarding it: you suggest that the PDF of the President's long form certificate might have been a DjVU file. As one of the links in my previous comment explained, a very similar process is used in creating PDFs directly from scanned color originals, without the PDF having to be in DjVU format at any point.

  9. Used by whom? Doesn't the procedure/program have a name at all? Of course it has. You just don't know anything about it.

    The search for the approximate clones to compress is known as JBIG2 and JB2 included in DjVu is arguably the only available alternative doing the same thing.

  10. Proof is in the embossed seal. The AP version lacks it , the White House version does not. No way Hawaii releases 2 versions, one embossed and one not. End of Story!

    "The AP version is a b/w scan and it didn't pick up the seal or the security paper."

    B.S. The paper is so thin the scan picks up the short form on another sheet behind the BC, but not one hint of the seal? Right. Oh, and the scan picks up some security background on the side but not the rest? Double Right.

  11. The leading premise of this is just sad. There is nothing magical about a subpoena. If I file a suit against you asserting that you were responsible for Kennedy's assassination I can get a clerk to sign off on a subpoena for all records associated with your activity for November 1963 in Dallas. That just means that I had the money to pay the fee.

    Orly Taitz is a nut. You should be embarrassed to be associating yourself with her but I have finally come to the conclusion that for a subset of conservatives there is no such thing as embarrassment if something will advance their agenda.

    Love your writings on physics but I shake my head when you wander off into this kind of stuff.