Laser printer attribution is an increasing problem with several applications, such as pointing out the ownership of crime proofs and authentication of printed documents. However, most of the existing methods are limited by modeling assumptions about printing artifacts.
In this paper, the authors explore solutions able to learn discriminant-printing patterns directly from the available data during an investigation, without any further feature engineering, proposing the first approach based on deep learning to laser printer attribution. Experimental results show that the proposed method is robust to noisy data and outperforms existing counterparts in the literature for this problem.
In summary, the main contributions of this paper are:
1) The design and development of an ad-hoc CNN for laser printer attribution based on the analysis of small patches representing text characters;
2) The use of CNNs on multiple representations of the same character to learn complementary features that are fused together for an increased recognition accuracy;
3) The use of a late-fusion paradigm to merge results coming from the analysis of different characters within the same document. In this way, each character is classified separately, and individual results contribute to the final document label.
Ferreira, Anselmo, et al. “Data-Driven Feature Characterization Techniques for Laser Printer Attribution.” IEEE Transactions on Information Forensics and Security (2017). doi: 10.1109/TIFS.2017.2692722