Friday, August 08, 2008

The great birth certificate "mystery": A challenge

No-one has refuted my smack-down of Techdude, an alleged "expert" in computer graphics who claims to have found the name of Barack Obama's sister hidden "between the pixels" on the much-debated certificate of birth. So I have decided to issue a put-up-or-shut-up challenge that should settle the matter once and for all.

To explain what this controversy is all about, let me take you back to the year 61 A.D. At that time, someone hammered bronze letters onto the east pediment of the Parthenon. The letters spelled "NERO IS LIKE TOTALLY THE COOLEST EMPEROR EVAH," or words to that effect. (I have translated the Latin rather loosely.) Seven years later, the unpopular Nero got "axed" to leave by his own guards, and those bronze letters were pried off of Athena's home.

But the holes remained.

They puzzled scholars for many years, until finally someone doped out what the inscription must have been. Only certain letters could match that particular pattern of holes.

In essence, Techdude claims to have done the same thing with the Obama "certificate of live birth," or COLB. (Here.)

The COLB consists (mostly) of black lettering on a pale green/white herringbone pattern. The pattern was on the "blank" paper to begin with. That patterned "blank" was shoved into a printer somewhere in the Aloha State, and the data was printed in black ink.

Are we clear on how the thing was originally made? Good.

Now, Techdude says that the alleged Obama COLB is actually a forgery. He says that the forger began with a COLB belonging to Maya Sotero, Barack Obama's half-sister.

The forger (we are led to believe) used a computer graphics program such as Photoshop to lift up a bit of the pale green/white herringbone pattern. This bit was taken from a part of the document that had no writing on it. The pattern was positioned over a piece of text that the forger wanted to cover up. The patterned overlay hovered over the text on a separate layer (which is one of the cool things you can do in Photoshop). The forger set this layer to "lighten," thereby erasing everything darker on the layer beneath -- such as that dark ink.

The result? After the layers were flattened together, the alleged forger had a nice blank patterned page, onto which he could place new writing.

Techdude claims that when the original writing was erased in this fashion, a few pixels remained. He says that one can recapture the hidden writing by studying those pixels. That's how he knows that the COLB was originally Maya's.

Sort of like studying those holes in the Parthenon. Ya dig?

Here's the rub: Everyone agrees that the holes in the east pediment are actually there. You can even find pictures of 'em on the web. But no-one has demonstrated that those "stray" pixels actually exist in the COLB.

So I decided to reverse engineer the thing. What I wanted to find out was this: If one were to follow in the footsteps of Techdude's imagined forger -- if one were to erase writing on the COLB using the method he suggests -- will ghostly hints of the original lettering be left behind? Will there be any stray pixels? Will (so to speak) any unsightly holes mar the east pediment?

Here's how I decided to answer that question.

First, I lifted out a chunk of the background pattern (using Photoshop, natch). That chunk is the greenish rectangle you see above. (Throughout this essay, click on the images to get enlarged, clearer versions.)

Next, I put that strip of light green herringbone right over the name "BARACK HUSSEIN OBAMA II." Yow! Look what happened!



You should be able to see the ghostly white outline around Barack's name. It's pretty damned obvious.

Is Techdude vindicated? No. If Maya's name had been that obvious, everyone would have noticed it right away.

So what went wrong? Why does that white outline exist?

When the COLB was posted to Daily Kos, it underwent heavy jpg compression. Why? Because it is a large image. Without compression, it would take forever to load up on your 'puter. Most images you see on the web are heavily compressed.

Compression introduces strange artifacts, especially in areas of high contrast, like black letters on a light background. Stray white and dark pixels will surround such letters. When I set the top-most pattern layer to "lighten," those white pixels beneath it were unaffected.

If the overlay were set to "normal" -- which is how I would have dunnit, frankly -- all of those stray pixels would be eradicated. So why wouldn't the forger do it my way, the safer way? Ya got me!

Now, when the proposed forger went to work, he presumably used Maya's actual COLB. Those black letters would have NO stray pixels. The result? This:



I saved this demonstration image at a high-quality setting, which means that the file is rather large. It's still a jpg, but it's a good jpg. There should be few weird artifacts around "BARACK," which I typed in using a 9 pt Arial font.

That's what you see in the field marked with a pink 1.

In field 2, you see "BARACK" covered by that same strip of overlaid pattern. Even though that pattern is on a layer set to "lighten" -- guess what? Every single pixel of the underlying "BARACK" is gone.

And I very much doubt that Techdude's much-vaunted forensic expertise will ever tease even one of those pixels back into visibility.

All of which brings us to the Great Techdude Challenge.

I have used what I understand to be the same techniques employed by the imagined forger. I have replicated what seems to be Techdude's notion of how the job was done.

I have typed a Shakespearean quote onto a blank area of the COLB, just above the place where you can read the words "CHILD'S NAME." I used 9 pt black Arial letters.

Then I covered up that quote, using the same techniques ascribed by Techdude to that supposed forger.

If Techdude's forensic methods actually work, then he should be able to tell us all what that quote is. Maybe you can uncover the quote in the image below. (Click to enlarge.) Play with the contrast slider! Change the hues! Use every filter you can think of! Brush up your Shakespeare -- start quoting him now!


The quote is quite obscure, so don't bother trying to guess it.

This test is perfectly fair. I have hidden this text using the same method allegedly used to hide Maya's name.

If the Dude now wants to argue that a different method was used, let him describe that method -- and let him tell us why a forger would prefer that method. I will prepare a similar test using any method he proposes.

C'mon, Techdude.

If you could find Maya's name in the COLB, you also should be able to divulge the secret Shakespearean text. The exact same methods should work in this case -- if they work at all, which I don't think they do.

In this instance, there are no holes in the east pediment of the Parthenon.

Why do I continue to challenge the Dude's work? Because Larry Johnson continues to promote it, online and on the radio. The PUMA movement is still a fairly delicate thing, and impressionable folks can be easily misled by this pseudo-scientific bullshit.

Frankly -- and at the risk of sounding paranoid -- I am starting to wonder whether Techdude is operating in good faith.

Oh -- and if folks out there want to question my good faith: Fine. Anyone else can put together an exactly similar experiment for the Dude. I have no problem with that, as long as the same rules are followed, and everyone is on the up-and-up. (If, as I suspect, Techdude is a bit of a con artist, then I naturally worry about the use of confederates.)

11 comments:

Anonymous said...

Rove?

Anonymous said...

thanks for this...

I found techdude's explanation of 'proof' incomprehensible myself -- the idea that one can 'prove' anything using stray pixels from a jpg file seemed rather bizarre to begin with.

That being said, the question is whether he's acting in bad faith or not. People see what they want to see (I mean, isn't that what the Obama campaign itself is all about?) and the fact that techdude 'found' something that probably isn't there doesn't mean he's acting in bad faith -- rather that he's human.

Of course, this really could be a 'ratfuck' operation -- as someone who did research on Bush's military records, I'm fully aware of how discrediting a prominent critic or evidence works to discredit the larger critique.

Twilight said...

If your doubts about Techdude are correct, what about the other expert, Ron Polarik ? His version of how a forgery was produced is different, and easier to follow.

He doesn't claim to have found Maya's (or anybody's) name, just that the certificate displayed by Kos and Fight the Smears has been tampered with, says he can identify the edges where a piece of blank background was added.

He commented in the last thread on this topic at Texas Darlin's blog.

Ray said...

Can I prentend to be Techdude's forger?

Here is a report of my first 2 attempts:


Ok, I started with a scan of his sister's COLB, which is a high quality TIFF file, and I wanted to leave tell-tale signs of her name, so I carefully covered it with blocks of background - but only JUST enough to hide the text and no more. I planned to leave in the artifacts on the perimeter of the name, but I found that there WERE none! Tiff files are not bastardized by compression and don't have artifacts around text.

I decided to convert the good Tiff file into JPEG format so I'd have the artifacts I needed to incriminate myself. That worked ok, so I tried again by covering ONLY the text and nothing beyond it.

I ended up with a very thin "perimeter of artifacts" 6 pixels above and below the covered panel. They looked all green and off-white, but when I increased the contrast greatly I saw patterns with stronger colours that could be vaguely equated with the alpha characters that had been there if some dude went poking around my forgery later.

Once I had planted my "artifact evidence" at the scene of the crime, I then typed in Obama's name in exactly the same position as his sister's name so that (a) It would look genuine and (b) so I wouldn't cover up any of the incriminating artifacts. If I placed the text too high or too low (by a few pixels) I'd prevent some dude from seeing the tell-tale artifacts.

The last step was to Save the forged image, but the bloody JPEG compression kicked-in and wrecked the artifact evidence that I had so carefully left behind. The NEW artifacts created a blend above and below Obama's name about 6 pixels wide -- NOT from black to light green as it normally would if a sane person was doing the forgery, but black to a dirty-greyish green. What's more, my new image stuck out like the rear end of a male dog. It had a bloody rectangle around the forgery that may as well have been a neon light!

As if THAT wasn't bad enough! To my horror I discovered that when I looked closely at the image I could see the rectangle of munged artifacts going the full length of his sister's name instead of stopping at the end of HIS name, so I had to measure the length of it and go back and cover that excess area with the background pattern.

Next attempt to follow

John said...

Well, I for one can't wait to see "TechDude" display his prowess for real. If he can't handle your test, he's a phony.

John
SluggoJD

Anonymous said...

If the Dude now wants to argue that a different method was used, let him describe that method -- and let him tell us why a forger would prefer that method.

Actually he put forth two possibilities over on his original Atlas Shrugs piece:

There are two obvious scenarios used to create the image that can be ascertained from evidence. Either a real COLB was scanned into Photoshop and digitally edited or a real COLB was first scanned to obtain the graphic layout then blanked by soaking the document in solvent to remove the toner.

se

Joseph Cannon said...

se: The second option is really stupid, unless the idea was to create a physical copy. But even then, you still have to remove the old text, put "virgin" pattern in its place, and then type in new text. So it all comes to the same thing.

I'm really starting to think that Techdude is spewing babble to impress the rubes.

Ray said...

One thing that has not been emphasized enough is that the "magic artifacts" need to be in a very narrow band between 1 and 8 pixels deep - above or below the text. All others are totally irrelevant.

Anonymous said...

That's some crackerjack forensic discourse, Mr. Cannon. As with the Parthenon evidence, but for rubes and plug-n-play simple folk, I'd like to describe a comparable phenomenon:

Throughout the Southeast you find Waffle Houses, there's a Waffle House everywhere. You also see places called Huddle House, Omelet Shop, and Omlet Shoppe. These others are hardly different from the (original) Waffle House - same menus, same interiors, same smell. The overwhelming impression is that these other restaurants are copies, knock-offs, barely legal forgeries of the original Waffle House. In fact, these imitations bought or leased old Waffle House restaurants and changed the name of the place, replacing the letters in 'Waffle House' on the outdoor tile signs with letters and words that would fit the existing tiles, including the blank space, which explains the Omelet Shop/Omlet Shoppe/Huddle House variants within the 12-tile character limit. Inside, it's still scattered, smothered, scattered and smothered, and so forth. No french-fries at Waffle House, but some of the others offer them.

Padraig

Citizen K said...

How big of a difference does the type of paper make in these experiments. In my line of work, we talk plain paper or copy paper, safety or security paper on which we print debentures, notes, bonds, etc., certificate paper, company stationery, in other words, different finishes. It would seem that type on copy paper and type on security paper are two different worlds. Are the differences inconsequential?

Ray said...

The paper for COLB's would have to be reasonably suited to laser printers.

I couldn't imagine the paper type or surface being an issue when it is only being scanned at a relatively course 300 ppi.

Ray