DDFed combines fully homomorphic encryption with cosine-similarity filtering so a federated learning server can aggregate encrypted updates and reject poisoned ones.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Dual Defense: Enhancing Privacy and Mitigating Poisoning Attacks in Federated Learning
DDFed combines fully homomorphic encryption with cosine-similarity filtering so a federated learning server can aggregate encrypted updates and reject poisoned ones.