This study explores the vulnerability of the Federated Learning (FL) model where a portion of clients participating in the FL process is under the control of adversaries who don’t have access to the training data but can access the training model and its parameters.