Trainers can plant statistically undetectable white-box backdoors in constrained DNNs that give exponential advantage for invariance-based adversarial examples, while outsiders cannot find any in poly-time under lattice assumptions.
Indistinguishability Obfuscation from
1 Pith paper cite this work, alongside 14 external citations. Polarity classification is still indexing.
1
Pith paper citing it
14
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Statistically Undetectable Backdoors in Deep Neural Networks
Trainers can plant statistically undetectable white-box backdoors in constrained DNNs that give exponential advantage for invariance-based adversarial examples, while outsiders cannot find any in poly-time under lattice assumptions.