Skip to main content

Table 1 Quantitative comparison of IGIMS and other two SoA algorithms in Single scenario

From: Iterative and mixed-spaces image gradient inversion attack in federated learning

Metric

CIFAR-10

FFHQ

ImageNet

 

IGIMS

GIAS

IG

IGIMS

GIAS

IG

IGIMS

GIAS

IG

MSE \(\downarrow\)

0.0030

0.0043

0.0065

0.0041

0.0056

0.0073

0.0217

0.0246

0.0420

PSNR \(\uparrow\)

26.33

24.41

22.88

25.00

23.27

21.98

16.68

15.86

14.06

SSIM \(\uparrow\)

0.8924

0.8602

0.8176

0.8737

0.8521

0.8001

0.4727

0.4274

0.2313

LPIPS \(\downarrow\)

0.0069

0.0092

0.0134

0.0073

0.0129

0.0166

0.4665

0.5510

0.7974

  1. We highlight the best performances in bold