Skip to main content

Table 6 Realistic performance of the individual detectors and fusion approaches with the adaptive decision thresholds and fusion weights based on the estimated EER 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 BPCER APCER HTER BPCER APCER HTER
DArXivMC Combined 30.08% 0.01% 15.04% 20.79% 0.89% 10.84%
DArXivNaive   19.21% 0.03% 9.62% 20.34% 0.50% 10.42%
DBIOSIGMC   39.76% 0.35% 20.05% 37.03% 4.51% 20.77%
DBIOSIGNaive   34.18% 0.65% 17.41% 33.76% 3.29% 18.52%
Dkeypoints   47.95% 1.15% 24.55% 73.17% 0.26% 36.72%
FM   26.81% 0.01% 13.41% 29.14% 0.42% 14.78%
FWLC   0.60% 10.87% 5.73% 0.30% 46.48% 23.39%
FDST   25.17% 0.02% 12.59% 33.53% 0.01% 16.77%
FLR   14.00% 0.09% 7.04% 14.31% 1.90% 8.10%
DArXivMC Complete 30.08% 0.00% 15.04% 20.79% 0.63% 10.71%
DArXivNaive   19.21% 0.02% 9.61% 20.34% 0.35% 10.34%
DBIOSIGMC   39.76% 0.38% 20.07% 37.03% 3.45% 20.24%
DBIOSIGNaive   34.18% 0.44% 17.31% 33.76% 2.46% 18.11%
Dkeypoints   47.95% 1.13% 24.54% 73.17% 0.09% 36.63%
FM   26.81% 0.01% 13.41% 29.14% 0.23% 14.68%
FWLC   0.60% 10.30% 5.45% 0.30% 41.15% 20.72%
FDST   25.17% 0.02% 12.59% 33.53% 0.02% 16.77%
FLR   14.00% 0.05% 7.02% 14.31% 1.30% 7.80%
DArXivMC Splicing 30.08% 0.03% 15.05% 20.79% 1.38% 11.08%
DArXivNaive   19.21% 0.05% 9.63% 20.34% 0.64% 10.49%
DBIOSIGMC   39.76% 1.11% 20.44% 37.03% 8.60% 22.82%
DBIOSIGNaive   34.18% 1.07% 17.62% 33.76% 4.35% 19.05%
Dkeypoints   47.95% 0.78% 24.36% 73.17% 0.25% 36.71%
FM   26.81% 0.01% 13.41% 29.14% 0.65% 14.89%
FWLC   0.60% 17.18% 8.89% 0.30% 56.84% 28.57%
FDST   25.17% 0.01% 12.59% 33.53% 0.03% 16.78%
FLR   14.00% 0.26% 7.13% 14.31% 3.53% 8.92%