Skip to main content

Table 4 Theoretical performance of the individual detectors with the combined LondonDB/UtrechtDB/Alabama datasets (best result per morph type marked in bold)

From: Potential advantages and limitations of using information fusion in media forensics—a discussion on the example of detecting face morphing attacks

   SC1 SC2
Detector Morph type EER τadaptive EER τadaptive
DArXivMC Combined 3.95% 0.528241 8.27% 0.417467
DArXivNaive   1.94% 0.594687 4.96% 0.499938
DBIOSIGMC   9.75% 0.617098 16.31% 0.561516
DBIOSIGNaive   7.74% 0.558175 13.56% 0.478364
Dkeypoints   12.65% 0.971509 19.08% 0.990942
DArXivMC Complete 4.00% 0.526729 7.75% 0.424468
DArXivNaive   1.82% 0.600357 4.06% 0.507648
DBIOSIGMC   9.75% 0.616997 14.98% 0.565297
DBIOSIGNaive   7.45% 0.563476 12.07% 0.496052
keypoints   12.43% 0.972011 19.53% 0.990758
DArXivMC Splicing 4.99% 0.501098 9.37% 0.406828
DArXivNaive   2.76% 0.566983 5.37% 0.492199
DBIOSIGMC   12.67% 0.594189 20.64% 0.541497
DBIOSIGNaive   9.08% 0.54235 14.84% 0.470685
Dkeypoints   11.09% 0.976876 19.08% 0.990933