In April of 2003, a stunning image appeared on the front page of the Los Angeles Times showing a British soldier stationed in Basra gesturing, gun drawn, toward a man carrying a child. It turned out that the image wasn't real. Two photos taken in quick succession had been combined by photographer Brian Walski in order to "improve the composition." He was quickly fired.
The manipulation was caught by an eagle-eyed employee at the Hartford Courant newspaper who viewed the image in Photoshop at around 600-times magnification.
Using Photoshop is the most common method for detecting such fakes, but it depends on tenacity and the fallible human eye. Computers are much more thorough. Recently, automated programs have been developed that can sift through digital images and look for suspicious stuff.
One of the oldest computerized methods consists of inserting a digital "watermark" into a photo, much like the ghostly logos embossed in the fabric of good paper.
Digimarc, a company from Oregon, provides services that allow subscribers to fix watermarks into digital images so that changes become immediately obvious. The company focuses on corporate copyright protection, though, rather than ethical breeches like the one attempted by Walski.
Hany Farid, a computer science professor at Dartmouth College, can detect fraud without the help of watermarks. Over the past few years he's developed a statistical algorithm that sifts through the pixels themselves for evidence of digital manipulation.
In much the same way that art historians analyze the brushstrokes in a painting to bust a high-profile forger, Farid's program looks for suspicious clumps of pixels amidst the noise of background information. For instance, when something is removed from a photo, or chunks are enlarged or reduced, the pixels in question look identical to their neighbours - a statistical impossibility.
Farid has been approached by everyone from UFO hunters to Brazilian supermodels interested in verifying images. The program is currently very complicated and is used only by professionals in the FBI and large media organizations. However, Farid plans to make a public version available within months.
This issue of digital manipulation is particularly hot in the world of science, ever since faked images from disgraced South Korean scientist Hwang Woo-suk graced the pages of Nature and Science.
The images were debunked by a method used by the Journal of Cell Biology, which has followed a policy of searching for tweaked images since 2002. Since then, it's found that 25 per cent of the papers submitted violate the journal's guidelines, ranging from simply cleaning up the background of a grainy photo to actually removing undesirable blobs from the depths of a petri dish.
The most radical method for tracking digital imagery comes from Jessica Fridrich at the State University of New York at Binghamton. Her technique traces photos to the cameras that took them, much as bullets can be traced back to individual guns.
Her proposal, though, would also take a picture of the eye looking through the camera and invisibly embed a version of the iris into the background of the photo. According to Fridrich, this would provide a much-needed link between the photographer, the camera and the image, an idea that has privacy advocates squirming.
In the meantime, photographer Brian Walski's website was temporarily shut down, and he's been quoted as saying, "I went from the front line for the greatest newspaper in the world, and now I have nothing. No cameras, no car, nothing."
It seems the best discouragement for unethical manipulation is still good old public pressure.