machine learning

Compression Footprints as Security Signals for Model-Poisoning Defense in Federated Learning