Single Fragment Forensic Coding from Discrepancy Theory

Abstract: Three-dimensional (3D) printing's accessibility enables rapid manufacturing but also poses security risks, such as the unauthorized production of untraceable firearms and prohibited items. To ensure traceability and accountability, embedding unique identifiers within printed objects is essential, in order to assist forensic investigation of illicit use. This paper models data embedding in 3D printing using principles from error-correcting codes, aiming to recover embedded information from partial or altered fragments of the object. Previous works embedded one-dimensional data (i.e., a vector) inside the object, and required almost all fragments of the object for successful decoding. In this work, we study a problem setting in which only one sufficiently large fragment of the object is available for decoding. We first show that for one-dimensional embedded information the problem can be easily solved using existing tools. Then, we introduce novel encoding schemes for two-dimensional information (i.e., a matrix), and three-dimensional information (i.e., a cube) which enable the information to be decoded from any sufficiently large rectangle-shaped or cuboid-shaped fragment. Lastly, we introduce a code that is also capable of correcting bit-flip errors, using techniques from recently proposed codes for DNA storage. Our codes operate at non-vanishing rates, and involve concepts from discrepancy theory called Van der Corput sets and Halton-Hammersely sets in novel ways.