Figure 1.

Schematic representation of federated learning combined with privacy-enhancing technologies (PETs). (1) Training of local models and implementation of PETs. (2) Secure transfer of local model parameters to a coordinator. (3A) Decryption of local models. (3B) Aggregation of local models. (4) Returning the global model to the participants. Note that these steps are often performed iteratively until model convergence (adapted from Huang et al [13]).