Research questionHow can federated learning reject poisoned client updates without server validation data or participant history?Malicious clients can submit updates that manipulate or degrade the global model. The server may lack trusted validation data and reliable prior knowledge about participating clients.