[> Most fingerprint systems convert the fingerprint image into [> what's called [> a template. This is a numeric representation, but [> comparision between [> two templates is not as simple as "==". Different portions of the [> template represent different minutae on the fingerprint, and [> an actual [> feature matching algorithm still needs to be used. Thus, we [> cannot hash [> these templates because there is no way to perform matching on the [> template hashes. There has been some interesting work done with facial biometrics that leverages the predictability of facial features (most eyes look similar, noses have a similar basic shape etc) to allow the generation of a 'false' facial image that will match to the template of a chosen target template. This iterates the placement and subtle variations in the faked facial image until this fake image results in an acceptable match through the same template generation process. The image doesn't need to be exactly the same (and the example images I saw were barely "human" to the naked eye) - merely that theey result in a similar template as the true image. I suspect the same process of iteratively varying whorls, ridges etc could, and would, result in a fales fingerprint that results in an acceptable template match to the real finger. Has anyone researched this with fingerprints? Lyal
