List your company
ResearchFormerly WizWand

Model Inversion Attack on VGGFace2

93.88Top-1 AccPPA
-1.633623.163247.9672.7568Mar 20, 2025
Updated 8mo ago

Evaluation results

MethodLinks
PPADefense=BiDO, Hyperparameters=0.006, 0.06, Target Model Accuracy=92.20, Target Backbone=MobileNetV22025.0393.8895.92——
PPADefense=MID, Hyperparameters=0.005, Target Model Accuracy=91.31, Target Backbone=MobileNetV22025.0393.88100——
PPADefense=LS, Hyperparameters=-0.001, Target Model Accuracy=92.40, Target Backbone=MobileNetV22025.0387.7695.92——
PPADefense=MID, Hyperparameters=0.005, Target Model Accuracy=89.50, Target Backbone=Swin Transformer2025.0383.6793.87——
PPADefense=TL, Hyperparameters=Block 4, Target Model Accuracy=93.82, Target Backbone=MobileNetV22025.0381.6395.92——
SMILE (InceptionV1*)Target Model=ResNet50, Image Prior=FFHQ, Surrogate Model=InceptionV1, Initialization=Pre-trained2025.0368.7193.2262.56237.75
PPADefense=LS, Hyperparameters=-0.0005, Target Model Accuracy=92.48, Target Backbone=Swin Transformer2025.0365.3177.55——
Mirror-wTarget Model=ResNet50, Image Prior=CelebA2025.0363.2779.59313.33280.97
PPADefense=BiDO, Hyperparameters=0.03, 0.3, Target Model Accuracy=91.57, Target Backbone=MobileNetV22025.0355.1175.51——
SMILE (InceptionV1*)Target Model=ResNet50, Image Prior=CelebA, Surrogate Model=InceptionV1, Initialization=Pre-trained2025.0344.2261.9339.16306.98
PPADefense=TL, Hyperparameters=Block 3, Target Model Accuracy=91.98, Target Backbone=Swin Transformer2025.0334.6959.18——
SMILEDefense=LS, Hyperparameters=-0.001, Target Model Accuracy=92.40, Target Backbone=MobileNetV22025.0330.6138.78——
SMILEDefense=LS, Hyperparameters=-0.0005, Target Model Accuracy=92.48, Target Backbone=Swin Transformer2025.0324.4934.69——
SMILEDefense=BiDO, Hyperparameters=0.03, 0.3, Target Model Accuracy=91.57, Target Backbone=MobileNetV22025.0322.4936.73——
SMILEDefense=BiDO, Hyperparameters=0.006, 0.06, Target Model Accuracy=92.20, Target Backbone=MobileNetV22025.0322.4540.82——
SMILEDefense=TL, Hyperparameters=Block 4, Target Model Accuracy=93.82, Target Backbone=MobileNetV22025.0320.4136.73——
RLBMIDefense=MID, Hyperparameters=0.005, Target Model Accuracy=89.50, Target Backbone=Swin Transformer2025.0318.3624.48——
SMILEDefense=MID, Hyperparameters=0.005, Target Model Accuracy=91.31, Target Backbone=MobileNetV22025.0314.2926.53——
RLBMIDefense=TL, Hyperparameters=Block 4, Target Model Accuracy=93.82, Target Backbone=MobileNetV22025.0314.2816.32——
Mirror-bDefense=BiDO, Hyperparameters=0.006, 0.06, Target Model Accuracy=92.20, Target Backbone=MobileNetV22025.0312.2416.33——
Mirror-bDefense=LS, Hyperparameters=-0.0005, Target Model Accuracy=92.48, Target Backbone=Swin Transformer2025.0310.214.29——
SMILEDefense=TL, Hyperparameters=Block 3, Target Model Accuracy=91.98, Target Backbone=Swin Transformer2025.0310.224.49——
RLBMIDefense=BiDO, Hyperparameters=0.006, 0.06, Target Model Accuracy=92.20, Target Backbone=MobileNetV22025.038.1616.32——
RLBMIDefense=BiDO, Hyperparameters=0.03, 0.3, Target Model Accuracy=91.57, Target Backbone=MobileNetV22025.038.1622.44——
Mirror-bDefense=MID, Hyperparameters=0.005, Target Model Accuracy=91.31, Target Backbone=MobileNetV22025.038.1624.49——
Mirror-bDefense=LS, Hyperparameters=-0.001, Target Model Accuracy=92.40, Target Backbone=MobileNetV22025.038.1610.2——
RLBMIDefense=LS, Hyperparameters=-0.0005, Target Model Accuracy=92.48, Target Backbone=Swin Transformer2025.038.1612.24——
Mirror-bDefense=BiDO, Hyperparameters=0.03, 0.3, Target Model Accuracy=91.57, Target Backbone=MobileNetV22025.036.1214.29——
RLBMIDefense=MID, Hyperparameters=0.005, Target Model Accuracy=91.31, Target Backbone=MobileNetV22025.036.1222.44——
SMILEDefense=MID, Hyperparameters=0.005, Target Model Accuracy=89.50, Target Backbone=Swin Transformer2025.036.1210.2——
Mirror-bDefense=TL, Hyperparameters=Block 4, Target Model Accuracy=93.82, Target Backbone=MobileNetV22025.036.1210.2——
Mirror-bDefense=MID, Hyperparameters=0.005, Target Model Accuracy=89.50, Target Backbone=Swin Transformer2025.034.084.08——
RLBMIDefense=LS, Hyperparameters=-0.001, Target Model Accuracy=92.40, Target Backbone=MobileNetV22025.034.0814.28——
RLBMIDefense=TL, Hyperparameters=Block 3, Target Model Accuracy=91.98, Target Backbone=Swin Transformer2025.034.086.12——
Mirror-bDefense=TL, Hyperparameters=Block 3, Target Model Accuracy=91.98, Target Backbone=Swin Transformer2025.032.046.12——