A single-class GAN discriminator with a calibrated confidence threshold detected 100% of tested FGSM-poisoned MNIST images at epsilon around 0.2 to 0.3, with weaker dirty-label separation for some digit pairs.
https://github.com/nisharaichur/Fast-Gradient-Signed-Method-FGSM Zhao, B., & Lao, Y
1 Pith paper cite this work, alongside 18 external citations. Polarity classification is still indexing.
1
Pith paper citing it
18
external citations · OpenAlex
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Dirty and Clean-Label attack detection using GAN discriminators
A single-class GAN discriminator with a calibrated confidence threshold detected 100% of tested FGSM-poisoned MNIST images at epsilon around 0.2 to 0.3, with weaker dirty-label separation for some digit pairs.