The RECODist Anselmo Castelo Branco Ferreira has earned his Ph.D. degree with the thesis entitled “Multi-Analysis Techniques for Digital Image Forensics” supervised by Prof. Anderson Rocha, from RECOD, and co-supervised by Prof. Jefersson Alex dos Santos.
The abstract is:
The development of approaches to authenticating and pinpointing the source of questioned documents attracted the attention of the research community in recent years, mostly because of the huge amount of information available today to ordinary people, such as videos and images, which can be easily tampered with to in order to produce deceitful information. Moreover, printed materials are daily created to forge documents such as currency, contractual clauses, and also are used to distribute child pornography photos. Solutions proposed in the literature often explore different branches of research, such as image manipulation detection, source attribution, steganalysis, pornography detection among others, by investigating characteristic artifacts in these images called structural inconsistencies. However, to investigate a given artifact for a given digital image forensic application, normally a very different approach is created. In this thesis, we aim at proposing new digital image forensic algorithms focused on different applications, but based on a core idea, called the multi-analysis. This new approach to create digital image forensic algorithms takes into account the analysis of several scenarios for the input image, such as the analysis of multi-directionality, multiple perturbations, multiple-resolutions, among others. These investigative scenarios can be applied in any step of the image investigation, such as pre-processing, description and classification. We show through an extensive series of experiments that the proposed solutions for laser printer attribution, image filtering detection and copy-move detection are efficient when compared with their literature counterparts and the approach proposed in this thesis can be the root of several other digital image forensic algorithms in the future.
Some highlights of his obtained results are:
1. FERREIRA, ANSELMO; NAVARRO, LUIZ C. ; PINHEIRO, GIULIANO ; SANTOS, JEFERSSON A. DOS ; ROCHA, ANDERSON . Laser printer attribution: Exploring new features and beyond. Forensic Science International, v. 247, p. 105-125, 2015.
2. FERREIRA, A. C. B.; ROCHA, A. . A Multiscale and Multi-Perturbation Blind Forensic Technique For Median Detecting. In: XIX Iberoamerican Congress on Pattern Recognition, Springer, 2014. v. 8827. p. 302-310.