Holograms detect digital
fraud
June 25, 2005
Courtesy Institute of Physics
and World Science Staff
A new technique for detecting forged photographs will help newspapers and magazines check pictures that might have been doctored, physicists say.
The new technology would also prevent hackers from tampering with sensitive legal images, including fingerprint records and medical scans used as evidence in court, and help intelligence officers verify the authenticity of photos.
Developed by physicists Giuseppe Schirripa Spagnolo and colleagues University of Roma Tre in Rome, Italy, this research was published on June 21 in the Journal of Optics A, a research journal.
In the system, an image, such as a company logo, is added to a digital photograph as an invisible “watermark”. Any subsequent attempt to alter the content of the photograph also results in damage to the watermark. Forgery could be detected using a computer to extract the watermark and check it for damage.
In a forged picture, a computer could even identify the object or section which has been tampered with, the researchers maintained.
To ensure that only an authorised recipient can extract the watermark, which could otherwise be put in a fraudulent image to pass it off as genuine, the watermark would be encrypted, or converted into a code, before adding it to the picture. Thus only someone who knows a private key can reconstruct it. The encryption also makes it hard to detect whether an image has been watermarked at all.
Before it is added to the photograph, the encrypted mark is turned into a computer generated hologram, the researchers explained. A normal hologram, an image that appears to be three-dimensional, is created with a process that uses lasers. The computer-generated version is a simulation of the pattern of light waves recorded when a real hologram is made.
As with all holograms, a small part of the hologram contains enough information to recreate the entire image. This means only a small part of the watermarked image is needed to extract the watermark.
When the hologram is added to the image it replaces the image “noise”, which has been filtered out beforehand. Noise is random information in a picture that doesnˇ¦t contribute to the image that you see, and can be removed without damaging the picture.
The watermark information is uses the same part of the light spectrum as the noise, making it invisible to the human eye in the photograph. The watermark is now embedded in the digital image file, in a separate part of the spectrum to the picture information, making it easy for the recipient to isolate and extract it. If the watermark can’t be reconstructed using the private key it means that someone has destroyed the watermark by trying to modify the image.
Testing their technique, the team demonstrated how a hologram watermark can be used to find out which part of an image has been tampered with. They changed colours in certain areas of a watermarked image and divided it into 16 parts, extracting a watermark from each. The parts where the colour change had taken place showed a significantly damaged watermark indicating that they had been modified.
“We hope that this technique can be used to improve the reliability of photographs in the media” said Lorenzo
Cozella, co-author of the paper, “Digital cameras could be developed so that an invisible watermark is added when a picture is taken. A newspaper buying a photo from a freelancer could then check for a watermark to confirm that it hasnˇ¦t been tampered with to make it more newsworthy.”
The system could also protect databases of images that serve as evidence in court, for example fingerprint records or medical scans, which could be used in cases of alleged malpractice. All images in the database would be watermarked so they could be checked without having to refer to the original hard-copy. This would be especially useful for electronically stored information exposed to external users via the internet, and thus to fraud, the researchers said.
* * *