Reversible watermarking could thwart digital photo tricksters
By Ben Coxworth
July 7, 2010
In these days of PhotoShop and its brethren, it’s becoming almost impossible to tell whether or not an image has been digitally manipulated. While some ‘shopping is done simply for whimsical reasons (see picture above), the matter becomes a bit more serious when things such as military images are altered. Visible watermarks are sometimes overlaid on digital photos, but these permanently alter and obscure that copy of the picture. Recently, however, researchers in India came up with a system for verifying a photo’s authenticity, without altering it in any way.
Dakshinamurthi Sivakumar and Govindarajan Yamuna of Annamalai University, in Tamil Nadu, have developed a process that they call reversible watermarking. Using a relatively small amount of computer power, their system measures the parameters of every pixel in an image. That data is converted into a Hash Message Authentication Code (HMAC), wherein the original pixel values are stored as a key.
After a visible watermark has been added and the image has been sent, the recipient simply uses the embedded HMAC to extract the watermark and restore the original image. Even if some unsavory character had intercepted the image and manipulated it in a way that didn’t alter the appearance of the watermark, the HMAC would still not match the pixel value of the current image, so the recipient would know something was up.
One would assume that the HMAC is unalterable after being initially created, although those hackers can be a pretty crafty bunch.
The research was recently published in the International Journal of Signal and Imaging Systems Engineering.