Wednesday, January 25, 2006

Remember the arms race between the Drug Runners and the Police in Florida?

The NYT has an article in the Scence Times of January 24, 2006 about what the Journal of Cell Biology does and plans to do, to catch doctored photos in article submissions. (Here's a copy of the Times story that may stay on the web for awhile.) It's officially illegal to submit a photo to this journal if part of it has been altered, but it's okay to uniformly alter the whole image (make it brighter, say). They have techniques for detecting whether part of the picture was dimmed or erased, and they will soon have programs from Dartmouth steganographer Hani Farid to detect cases where part of the image seems to have a differently oriented light source. Harid's programs will also detect exact copies of parts of the image within the image, because a common way to erase something is to copy part of the background over it.

Near the end of the article they quote a doubtful editor, Emilie Marcus, who notes that the whole process of scientific journalism is based on trust. She asks "Why say 'We trust you, but not in this one domain?'" And she's definitely on to something. Which brings us to the arms race between the drug runners and the police in Florida. There was a time when the police needed the fastest possible boats to catch the runners. And the runners needed even faster boats to escape. And they both got their boats from the same designer, who did very well creating ever faster boats. (I didn't fact-check that story, hope it's not an urban myth.)
Any good applied math programmer could write programs to smooth, adjust or roughen images in order to conceal exactly what Hani Farid's math will look for. And then someone can write software to detect that type of image manipulation, someone can write software to conceal it, and so on. We're looking at another war that will definitely profit the programmers, and Cell would be clever to try to make Trust work.

No comments: