A simple technique used in 12th-century Italy to authenticate high-quality paper has met 21st-century researchers trying to protect the authentication and copyright of digital files through digital watermarking.

It's the digital equivalent to holding a photo, video or music file up to the light to see whether or not it is authentic or has been altered. P.K. Rajan, chair of Electrical and Computer Engineering, has taken up the challenge of refining the technique so that copyrighted digital photos, videos or music files can be more effectively watermarked without distortion and so that files can be easily identified as authentic.

"Copyright and authentication are huge issues in this digital age where we produce and transmit so many photos, sound files and other works digitally," says Rajan. "We want to improve the level of protection of these works because there are potential serious consequences when digital files are used without authorization or authenticity is compromised.

"For instance, if a forensic specialist takes a digital photo of a murder scene that is entered into evidence, jurors need some reassurance that the photo hasn't been modified in any way with software that's in common use. In this case, we need a watermark that allows acceptable manipulation, like resizing, without allowing alteration. Currently, digital photography is considered 'untrustworthy' since it is so easy to edit."

Rajan's example refers to one of the two categories of current problems addressed by the digital watermarking research community — robust watermarks for copyright security and fragile watermarks for authentication of images. Watermarking can be used for copyright security because it allows a hidden tag that says "this belongs to ..." to be inserted into the data in an invisible manner. The watermark is robust so that the watermark cannot be deleted without destroying the image itself.

"Removing a robust code is like removing an ink-filled security device on a piece of clothing without the proper device," says Rajan. "You cannot remove either without destroying the product.

"On the other hand, fragile watermarks are designed to be broken when the file is altered so that you can detect if a file has been altered in a significant way," he says.

To be effective, a watermark must be embedded in the perceptually most significant portions of an image or in the most identifiable audio or video portions of a file, but it is difficult to embed a watermark in such important places in a way that does not distort the original image or sound. The watermark bits must be scattered in these most significant portions of the image or music so that they cannot be seen or changed.

Rajan and undergraduate students David Cook and Jameson Porter, who were supported by an NSF Research Experience for Undergraduates grant, worked on this project last summer. He and Cook presently focus on developing a new layered watermarking algorithm that improves the localization accuracy of the watermark, meaning you can more precisely detect which part of the image has been modified.

They want to combine this high accuracy of detecting the watermark with a public-key encryption system. Public-key encryption is used heavily in "digital signatures" that verify who sent the file; the person receiving it does not have to have any special software or equipment to read the signature, but he or she cannot alter it.

"Simply, we want to retain the ability to detect tiny changes while also retaining the security offered by public-key encryption," explains Rajan. "Public-key encryption is important since it allows the decryption of the data without having to publish the encrypting key. That is, the security of the key need not be compromised to allow the authentication of only one image."