Modeling Attacks on Photo-ID Documents and Applying Media Forensics for the Detection of Facial Morphing

摘要: Since 2014, a novel approach to attack face image based person verification designated as morphing has been actively discussed in the biometric and media forensics communities. Up until that point, modern travel documents were considered be extremely hard forge or successfully manipulate. In case of template-targeting attacks like facial morphing, process becomes vulnerable, making it necessity design protection mechanisms. this paper, new modeling for is introduced. We start with life-cycle model photo-ID documents. extend by an editing history model, allowing precise description realizations foundation performing well training testing scenarios detectors. On basis these approaches, two different forensic detector are implemented evaluated. The feature space on idea blending operation pipeline causes reduction details. To quantify reduction, we adopt features OpenCV processing library, namely number SIFT, SURF, ORB, FAST AGAST keypoints region loss edge-information Canny Sobel edge operators. Our trained 2000 self-acquired authentic morphed images captured three camera types (Canon EOS 1200D, Nikon D 3300, Coolpix A100) tested from public database. Morphing detection accuracies decision tree classifier vary 81.3% 98% test scenarios.