Abstract: Device identification, the ability to discern the (separate) devices by which a document was produced and/or imaged, can be leveraged in the design of quality assurance (QA) systems as well as the practice of forensic analysis. It is shown that QA metrics associated with printed security markings provide a useful approach for performing multiple device identification, i.e., printer-scanner identification. While some previous methods have focused on properties of sensors to extract signatures from general image data, the proposed approach leverages the highly structured nature of color tile deterrents to predict device (combination) signatures based on a limited amount of information. Constraints introduced by the deterrent structure yield a relatively simple classification strategy with strong performance using a 10-dimensional feature vector. Sixteen printer-scanner combinations (composed from 4 printers and 4 scanners) are tested using this method. Results illustrate device signature prediction performance that is competitive with current state-of-the-art approaches based on physical models of the devices involved.

  author       = {Matthew D. Gaubatz and Steven J. Simske},
  url          = {http://www.hpl.hp.com/techreports/2009/HPL-2009-370.pdf},
  year         = {2009},
  pages        = {151--155},
  address      = {London, UK},
  title        = {Printer-scanner identification via analysis of structured security deterrents},
  booktitle    = {Proceedings of the 2009 First IEEE International Workshop on Information Forensics and Security},