Hear No Evil

Federated learning gradient leakage attacks are detectable, study finds

Researchers have analyzed how malicious servers can exploit federated learning to extract sensitive data from client updates. They found that while these attacks are theoretically concerning, they are often detectable a…