Co-Principal Investigators: Jan P. Allebach, Edward J. Delp, and George T. Chiu
We propose to develop two strategies for printer identification. The first strategy is passive. It involves characterizing the printer and finding intrinsic features in the printed output that are characteristic of that particular printer, model, or manufacturer's products. We call this the intrinsic signature. Developing the intrinsic signature requires an understanding and modeling of the printer mechanism, and the development of image analysis tools that are used for printer characterization during the signature development phase, and then later, for the actual detection of the signature in printed pages with arbitrary content.
The intrinsic signature is detected by scanning the printed pages with a high resolution drum scanner, and applying low-level image analysis routines to extract features. These features are processed with a soft classifier to yield likelihoods at each level of a decision tree that the document was printed with a device belonging to each particular class. At the highest level of the decision tree, likelihoods are provided for which of the two possible dominant printing technologies: electrophotography (commonly referred to as a laser printer) and inkjet was used. At the next level,likelihoods are generated for the candidate printer manufacturers, and so on. As we proceed down through the tree, we generate liklihoods regarding information that is more and more specific to the particular printer in question.
The second strategy is active. Here we embed an extrinsic signature in every printed page. This signature is generated by modulating the process parameters in the printer mechanism to encode identifying information, such as the printer serial number and date of printing, in every printed page. To detect the extrinsic signature, we again scan the printed pages, and process them using image analysis techniques; but in this case, our goal is to decode the signature to extract the information embedded in it. Development of the methodology for extrinsic signature embedding will build directly on our work with intrinsic signatures. We will use our knowledge of the printer mechanism models and the results of the printer characterization to determine the printer process parameters that can be modulated to encode the desired identifying information. The modulation of these parameters will require modification to the actual printer mechanism.
A distinguishing feature of the proposed effort will be the development of an undergraduate project course that will be associated with the research. In this course, students will learn about printing technologies and the application of electrical and mechanical engineering theory from their core courses to analysis and modeling of printing systems. They will also learn about image processing and decision theory; and they will see how all these tools can be applied to the solution of practical real-world problems.
