List your company

Targeted Transfer Attack on ImageNet (val)

100Attack Success RatecAdv
16.38438.09259.881.508Oct 5, 2022

Evaluation results

MethodLinks
cAdvSubstitute Model=ViT-S, Target Model=ViT-S, Attack Type=White-box2022.10100————————————————
cAdvSubstitute Model=XCiT-N12, Target Model=XCiT-N12, Attack Type=White-box2022.10100————————————————
ACESubstitute Model=ViT-S, Target Model=ViT-S, Attack Type=White-box2022.1099.7————————————————
ColorFoolSubstitute Model=ViT-S, Target Model=ViT-S, Attack Type=White-box2022.1099.2————————————————
ACESubstitute Model=XCiT-N12, Target Model=XCiT-N12, Attack Type=White-box2022.1098.7————————————————
SAESubstitute Model=ViT-S, Target Model=ViT-S, Target Architecture=Transformers, Attack Type=White-box2022.1098.2————————————————
ReColorAdvSubstitute Model=ViT-S, Target Model=ViT-S, Attack Type=White-box2022.1096.2————————————————
ReColorAdvSubstitute Model=XCiT-N12, Target Model=XCiT-N12, Attack Type=White-box2022.1095.9————————————————
NCFSubstitute Model=ViT-S, Target Model=ViT-S, Target Architecture=Transformers, Attack Type=White-box2022.1093.9————————————————
NCFSubstitute Model=XCiT-N12, Target Model=XCiT-N12, Target Architecture=Transformers, Attack Type=White-box2022.1089.1————————————————
SAESubstitute Model=XCiT-N12, Target Model=XCiT-N12, Attack Type=White-box2022.1086.5————————————————
ColorFoolSubstitute Model=XCiT-N12, Target Model=XCiT-N12, Attack Type=White-box2022.1085.9————————————————
NCFSubstitute Model=XCiT-N12, Target Model=Res-18, Target Architecture=CNNs, Attack Type=Black-box2022.1065.1————————————————
NCFSubstitute Model=XCiT-N12, Target Model=VGG-19, Target Architecture=CNNs, Attack Type=Black-box2022.1063.7————————————————
NCFSubstitute Model=ViT-S, Target Model=ViT-B, Target Architecture=Transformers, Attack Type=Black-box2022.1062.7————————————————
NCFSubstitute Model=XCiT-N12, Target Model=Mobile-v2, Target Architecture=CNNs, Attack Type=Black-box2022.1062.6————————————————
NCFSubstitute Model=ViT-S, Target Model=Res-18, Target Architecture=CNNs, Attack Type=Black-box2022.1057.5————————————————
SAESubstitute Model=ViT-S, Target Model=Res-18, Target Architecture=CNNs, Attack Type=Black-box2022.1049.9————————————————
SAESubstitute Model=ViT-S, Target Model=VGG-19, Target Architecture=CNNs, Attack Type=Black-box2022.1047.6————————————————
SAESubstitute Model=ViT-S, Target Model=Mobile-v2, Target Architecture=CNNs, Attack Type=Black-box2022.1046.5————————————————
SAESubstitute Model=ViT-S, Target Model=XCiT-N12, Target Architecture=Transformers, Attack Type=Black-box2022.1038.7————————————————
SAESubstitute Model=ViT-S, Target Model=ViT-B, Target Architecture=Transformers, Attack Type=Black-box2022.1037.5————————————————
SAESubstitute Model=ViT-S, Target Model=PiT-Ti, Target Architecture=Transformers, Attack Type=Black-box2022.1034————————————————
SAESubstitute Model=ViT-S, Target Model=DeiT-S, Target Architecture=Transformers, Attack Type=Black-box2022.1024.2————————————————
SAESubstitute Model=ViT-S, Target Model=Swin-T, Target Architecture=Transformers, Attack Type=Black-box2022.1019.6————————————————
Admix-RDIsource_model=RN-50, MI-TI=applied2023.05—20.974.2—42.759.486.8————80.798.756.126.734.158
CFM-RDIsource_model=RN-50, MI-TI=applied2023.05—51.184.7—65.581.590.3————88.498.478.84859.374.6
DIsource_model=RN-50, MI-TI=applied2023.05—5.762.5—9.228.272.3————56.698.929.34.59.937.7
FDA*Loss function=feature distribution attack, Source robustness (epsilon)=0, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—43.5———22.957.9——————————
LogitLoss function=logit loss, Source robustness (epsilon)=0, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—45.94015.343.622.977.960.339.63.90.4——————
LogitLoss function=logit loss, Source robustness (epsilon)=0.1, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—77.278.864.573.152.58481.668.933.411.2——————
LogitLoss function=logit loss, Source robustness (epsilon)=1, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—56.162.459.55547.26758.353.63626.7——————
ODIsource_model=RN-50, MI-TI=applied2023.05—43.878.3—65.967.387————77.197.67049.555.469.2
Po+TripLoss function=Poincaré distance + triplet loss, Source robustness (epsilon)=0, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—20.815.2102311.659.331.214.21.30.3——————
Po+TripLoss function=Poincaré distance + triplet loss, Source robustness (epsilon)=0.1, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—59.157.95356.539.278.472.645.111.43.3——————
Po+TripLoss function=Poincaré distance + triplet loss, Source robustness (epsilon)=1, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—48.554.460.249.539.962.653.3482212.4——————
RDIsource_model=RN-50, MI-TI=applied2023.05—13.165.4—30.746.681.3————71.89846.616.823.949.4
SI-RDIsource_model=RN-50, MI-TI=applied2023.05—29.570.5—56.456.288.9————79.898.866.237.943.662.8
VT-RDIsource_model=RN-50, MI-TI=applied2023.05—27.968.8—45.854.582.5————78.798.256.132.837.958.3
XentLoss function=cross-entropy, Source robustness (epsilon)=0, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—10.49.64.610.66.440.513.16.80.80.1——————
XentLoss function=cross-entropy, Source robustness (epsilon)=0.1, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—5459.445.850.832.178.86641.18.62.4——————
XentLoss function=cross-entropy, Source robustness (epsilon)=1, Source architecture=ResNet50, l_infinity norm constraint=16/2552021.06—60.469.366.658.246.769.961.356.929.919.9——————