In the near future, it could be a lot easier to see if those pictures of the person you've been talking to on Match.com have been retouched.
Adobe Systems' Advanced Technology lab is working on plug-ins for Photoshop that would detect whether a photo has been tampered with, according to an Adobe representative. So far, the company has two plug-ins that are in a fairly advanced stage of development. Adobe is working with Dartmouth professor Hany Farid, an expert in photo fraud detection.
One tool from Adobe, called Clone Tool Detector, determines whether a section in a picture, such as a patch of sand or a field of grass, has been recopied from another part of the picture. Last year, Reuters admitted that a photographer cloned a smoke plume in a shot of wartime Beirut. The tool can't ascertain with absolute certainty whether two items, say a pair of clouds or other images, have been cloned, but it will say that the two images are "improbably similar."
The other tool, informally known in the lab as Truth Dots, determines whether pixels are missing from a photo, a sign that the image has been cropped. Even if the images were greatly magnified, the human eye could not detect these.
The representative said that Adobe has not made a final determination of whether or when to add these tools to Photoshop, but added that Adobe is very interested in photo authentication, which has become a hot topic with the proliferation of blogs, photo sharing Web sites and online news.
Farid has worked with Adobe, the FBI and news agencies for several years on detecting photo and image fraud. Informally, he also gets a constant stream of requests from people asking whether the picture of an item on eBay is real, or from people who want the picture of their potential date on dating sites to be verified. (He and the researchers at the Dartmouth Image Science Group have also worked with art museums to determine the authenticity of certain masterpieces.)
More recently, Farid has also begun to develop tools that could sniff out video and audio forgeries, a much more difficult task.
Probability plays a significant role in fraud detection. The software essentially looks for anomalies and discontinuities in pictures and flags them. An underlying understanding of the hardware and software used by photographers also helps.
Different cameras from the same manufacturer operate under different JPEG quantization tables, which determine the rate at which a camera will drop data in compressing a photograph. Farid's group has come up with software for examining the quantization tables across different cameras.
Adobe Photoshop, meanwhile, has its own distinct quantization table. As a result, the software can tell whether a photo has been run through Photoshop or came from a source other than that claimed.
"I can't tell you the serial number of the camera, but I can tell you this did not come from a Canon PowerShot. It came from a Nikon," he said in an interview last fall. "You can also tell if it came through Photoshop. It won't tell you what happened to the image, but it tells you it did not directly come out of the camera."
In an e-mail interview today, Farid did not comment on the exact nature of what Adobe might include in Photoshop but said it would likely be less powerful than the software he provides to law enforcement agencies.
"The primary reason for this is that some of our tools still require a fair amount of expertise to use," he wrote.
Making more authentication tools will likely help consumers sniff out fraud, but it could also help people interested in concocting fraudulent or doctored pictures. Child pornographers regularly exploit photo doctoring tools to develop composite pictures that allow them to get around criminal statutes. The 2004 presidential campaign also showed how doctored photos--remember the fake photo of John Kerry and Jane Fonda sitting next to each other--can influence a campaign.
"'Educating fraudsters' is a real issue, but also an inevitable one in this type of game. I do believe that although we publish all of our results, the combination of techniques that we and others are developing will make it increasingly more difficult to create convincing forgeries," Farid wrote. "Whether we publish our results or not, the best forgers will always be able to create forgeries that go undetected. The average person, however, will likely not be able to circumvent all of our techniques."
...and will prolly last about as long as it takes for someone to cook up an EXIF data editor capable of breaking whatever encryption they toss at it (or simply rips the old one out and sticks a replacement EXIF header in...)
No, I don't think so, you don't need EXIFs at all for this. The software examines the pixels. Each camera produces noise, and the pattern of that noise, in addition to depending how it was compressed, will vary between different brands. After all, the EXIF is maintained, even though the photo can be manipulated. So, manipulation is only detected by looking at the actual image.
Good Photo editors could still "edit" without detection
As a graphic artist who's used PhotoShop from 1.0 to the current version, I doubt this "detection" will work if someone with enough experience does the editing. There's ways to get around this. If a photo is "edited" without using the cloning tool the technique won't work, will it? Printing an "edited" photo out and then rescanning (with a filter?) may be another way. And those two ideas are just off the top of my head.
Why even bother? I could edit a photo, print it out with a high definition printer and then take a picture of it...or project it...or any number of tricks
can be used as a tool to help make fraud harder to detect.
First, keep editing your photo until everything but the quantization tables checks as legit. Then save using a plugin that uses Canon quantization tables.
You would need to use original photos, or stock shots taken in raw mode. Double compression would be detectable. The only truly hard part would be writing the canon plugin.
Google creates an animated doodle that features a boy, a girl, Google's search engine, and a jump rope. But might there be darker, more analytical, more troubling interpretations to this tale?
The Silicon Valley online payments startup grew by 1,000 percent last year and is hopeful it can repeat that level of growth this year. To do that, it's had to move away from its early friends-and-family roots and embrace small businesses.
Chamtech's spray-on antenna uses a nano material to provide a low-power boost to antenna range. The wireless-in-a-can product may some day bring an end to unsightly cell towers.
EnerG2 opens a plant to make an engineered carbon that will improve performance of energy storage devices and make storage for start-stop hybrid cars less expensive.
I give it about 24 hours. Perhaps 48.
/P
The software examines the pixels. Each camera produces
noise, and the pattern of that noise, in addition to depending how
it was compressed, will vary between different brands.
After all, the EXIF is maintained, even though the photo
can be manipulated. So, manipulation is only detected by
looking at the actual image.
version, I doubt this "detection" will work if someone with enough
experience does the editing. There's ways to get around this. If a
photo is "edited" without using the cloning tool the technique won't
work, will it? Printing an "edited" photo out and then rescanning
(with a filter?) may be another way. And those two ideas are just
off the top of my head.
Now Adobe slips into the uniform of CSIs
You would need to use original photos, or stock shots taken in raw mode. Double compression would be detectable. The only truly hard part would be writing the canon plugin.