Marking digital files with a fingerprint of some kind – a copyright for instance – is crucial for protecting proprietary information. A music purchase can be marked with tracking information such as who bought it and when. A photo can be dated to prove to whom it belongs.
But for every form of encrypted information, there’s a way to undermine it. “Over the years, people have come up with a lot of creative ways to detect digital watermarks,” says Scott Craver, assistant professor of electrical and computer engineering at Binghamton University. Once detected, they can be broken. “I mostly work on evading those on the detection side.”
New methods of embedding watermarks are continually devised and new attacks on breaking those watermarks follow in short order. “It’s a cat-and-mouse game,” Craver says.
Detection methods use statistical sampling of pixels in a two-dimensional image (or voxels in a 3D image). The differences between natural and tampered data are subtle, Craver says, but with sophisticated computer programs, they can be detected.
Craver is working on designs that are “somewhat artificial” at the outset, so that watermarks are invisible to computer sampling programs. By purposefully making an innocuous background that generates ‘watermark-like statistics,’ a watermark can be embedded without registering a statistical difference.
Craver received a Presidential Early Career Award for Scientists and Engineers, the highest honor bestowed by the US government on outstanding scientists early in their careers.
He has been at Binghamton since the fall of 2004 and has found a stimulating research environment, working on information security with several other Binghamton researchers, including Jessica Fridrich, who works on multimedia forensics.
Learn more about
Research at Binghamton University
Other areas of research and discovery by the Research Foundation