A sparse column-wise perturbation plus isotropic Gaussian dither plants a backdoor in CNNs and ViTs that is as hard to detect as Sparse PCA under standard hardness assumptions.
BAN: Detecting backdoors activated by adversarial neuron noise
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Undetectable Backdoors in Model Parameters: Hiding Sparse Secrets in High Dimensions
A sparse column-wise perturbation plus isotropic Gaussian dither plants a backdoor in CNNs and ViTs that is as hard to detect as Sparse PCA under standard hardness assumptions.