Unretouched by Human Hand

The need to protect photographic images from being tampered with is finally being acknowledged

Like the Dutch boy with his finger in the dyke, the entertainment lobby has been vainly trying to stop the torrent of illegally downloaded digital movies and music ever since the demise of Napster and the emergence of other software for swapping video and audio files over the Internet. Copyright schemes for video and music abound—including everything from pay-per-view web video to Celine Dion compact discs that deliberately crash computers. However, there has been far less talk about new tactics to defend digitised photographs and other images from tampering, copying and redistribution.

That could change. A group from the University of Rochester, New York, and Xerox Corporation is quietly perfecting a procedure called “reversible data-hiding”. The technique verifies the integrity and authenticity of digital images, assuring users that they are viewing an untainted original, and notifying the creator if a photograph is misappropriated.

Protecting a digital photograph is trickier than safeguarding a digital video or music file. Both processes involve a programmer embedding encryption, copyright or authentication (“digital watermark”) code somewhere within the binary data that make up the image. Digital video or music files are more straightforward to padlock because of their large size. It requires a hefty amount of data to reproduce a video or a musical score. That makes it easier to add sections of code here and there without corrupting the overall quality of the file. A still image, by contrast, contains fewer bits (ie, binary digits of ones and zeros). With limited space, adding protective measures means sacrificing some of the image data, with a noticeable loss in picture quality. And for many purposes that is unacceptable.

Reversible data-hiding offers a clever solution. The embedding process begins by extracting and compressing some minor details of the image, explains Mehmet Celik of the University of Rochester. This packet of data (usually about four kilobytes of code) is then embedded within the file, and a digital watermark is inserted into the newly created space left behind by the compression process. When authorised users open the file, they see the watermark, which tells them that the image has not been manipulated or corrupted in some irreversible way. Software then automatically removes the watermark, decompresses the “hidden” packet of data, and rebuilds the image to its former specifications.

It sounds like a lot of trouble for one image. But as Mr Celik outlines the possible applications for reversible data-hiding, it becomes obvious why preserving image integrity is so critical. In a courtroom, for instance, a police officer might want to present photographic evidence taken at the crime scene with a digital camera. Usually, there is no way to know whether the images have been altered. With reversible data-hiding, however, photographic evidence could be watermarked and secured from tampering the moment the shutter is pressed.

The process improves on the idea of digital signatures, which require both the creator and the viewer to “unlock” the file with a unique “key”. This may be all but foolproof, but it takes up more room; and the signature could be stripped from the file when the image is opened and saved again. The data-hiding technique neatly fits within the image, and all the authentication tasks are localised, notes Guarav Sharma of Xerox. This does not prevent the image from being edited with software such as Photoshop, but it will tell authorised users whether the image is in its original state or has been retouched.

The United States army has expressed an interest in the data-hiding algorithms for combining extremely high-resolution reconnaissance photographs with authentication security. Eventually, Mr Sharma expects to see Xerox scanners and copiers with data-hiding functions built into them.