Last week, the state-run North Korean media released what they said were new photographs of Kim Jong-il in a bid to reassure the world that the leader, whose health has been the subject of much speculation, is alive and well. Within days questions were raised by British news agencies, the Times of London and the BBC, about the legitimacy of the photos, citing apparent inconsistencies within the images.
But were the photos faked?
We contacted digital-forensics expert Hany Farid, professor of computer science at Dartmouth College and author of a June Scientific American article on detecting fake images, for his take. In a phrase: More than likely not.
"What the BBC and the Times wrote can be explained," Farid told us. "It's not indicative of tampering." He says: "The BBC pointed to, I think, three or four things that they thought were indicative of tampering. And in the full-resolution image you can see that they're completely wrong." (Farid reviewed the photograph at the request of the Associated Press.)
One alleged manipulation was between the shadow cast by Kim's legs and the shadows cast by the soldiers flanking him—a discrepancy that Farid says can be resolved by accounting for a curvature of the background surface. "If you look closely at the back baseboard, which all the men are standing at, it's actually curved," he says. "So what [the news agencies] are assuming is that that backboard is straight. If that backboard is straight and there's only one light source, they're right, it's hard to explain that difference in the shadow.”
But assuming a curvature, the men would naturally cast different shadows. Farid calculates that only a few inches of background difference would suffice: "the sun is at such a grazing angle, so small differences make huge variations in the length of the shadow."
The BBC also pointed to "apparently mismatched pixels" around Kim's legs. But Farid says that the BBC "did this thing that is very dangerous, which is they zoom and they say, 'Oh look, it's a splice line around his feet,'" indicating that the leader may have been edited in. The problem with such an approach, he says, "is that's all JPEG compression artifacts, and if you actually do the same thing to anyone else's feet you see exactly the same artifacts." Image compression uses a sort of digital shorthand to reduce the size of the files, throwing out certain nuances in favor of approximations that can be somewhat choppy. "What that means is there's a quite a bit of color artifacts when you zoom in like that. So you completely expect those types of thing."
Another discrepancy, a discontinuity in a black line running horizontally behind Kim and the soldiers, could be a removable piece of the baseboard, perhaps to allow access to the bleachers above, Farid says, but he prefers not to make such "qualitative" assessments, relying instead on digital forensics.
"Quantitatively, I ran a number of forensics tools, and there's no cloning, there's no color-filtering artifacts, the lighting is completely consistent, you can explain the shadows," Farid says. "The image was edited, as all images are, because they all get cropped and contrast-enhanced, but other than that, there was just no signs of tampering anywhere."
Of course, that's not to say that North Korea is always up-front with the facts, nor is it to say that the image, if legitimate, proves anything about Kim's health. Even under the best of circumstances, Farid adds, "I never say an image is authentic; I say I can't find traces of tampering. It's a subtle but important distinction."
CREDIT: ASSOCIATED PRESS