Tag: model poisoning

Researchers have developed two federated learning algorithms, Bi-Models Training and Multi-Models Training, that remove a client's influence from a shared ...

Researchers at Donghua University have developed RFLDefense, a layer-wise defense framework that suppresses backdoor attacks in federated learning even when ...

A massive 641-run study shows that trust-weighted federated learning can keep a tiny on-device IoT intrusion detector accurate under non-IID ...

Researchers have demonstrated a switchable backdoor attack that injects dual tokens through the layers of Vision Transformers, achieving up to ...