FedRE allocates layer-wise local differential privacy noise based on gradients' sensitivity to client-specified privacy-sensitive regions, plus a perturbation-aware aggregation, for federated text tamper detection.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
FedRE: Robust and Effective Federated Learning with Privacy Preference
FedRE allocates layer-wise local differential privacy noise based on gradients' sensitivity to client-specified privacy-sensitive regions, plus a perturbation-aware aggregation, for federated text tamper detection.