A new report by the CITP describes a technique for identifying and authenticating paper documents using a desktop scanner and custom software, that starts with identifying “inhomogeneities in the substrate:”
Our method is an improvement over previous work because we measure the surface texture of a document without the requirement of expensive equipment. We utilize the unique fiber structure as identified and relied upon by Metois et al., Cowburn and Buchanan, and Zhu et al. but do so without modifying the document in any way. Our method allows documents to be fingerprinted before or after content is printed. In fact, fingerprinting and tracking using our system can begin during the paper manufacturing process. We have also developed methods for hiding the target feature vector through the use of a secure sketch. This means a potential counterfeiter cannot learn what features he needs to reproduce from the fingerprint alone but would need access to the original document to even attempt a forgery. (p. 2)
In that case?
Suppose an attacker has seen an original document and wants to create a second document that will pass verification as the original. The attacker will start with an arbitrary piece of paper which, by assumption, will have a very different feature vector from the original. He will then try to modify the target paper so that its feature vector is close to that of the original document. To do this, the attacker needs to make fine-grained modifications to the document’s surface normals. This might be done via lithography or photographic techniques, but these will be expensive and will probably require special equipment. The most effective, economical way to control the surface, we believe, is to print on the document. (pp. 9-10)
An attacker could also use the method to “de-anonymise” paper forms like a ballot or survey. It’s like CSI for paper nerds. (Honey, I’m looking in your direction.)