Federated and privacy-preserving learning.

Technologies

• Federated learning across sites Training shared models where data stays at each site and only updates move.
• Federated learning in 5G systems Network-supported federated learning between analytics functions and devices.
• Differential privacy Formal limits on what any output reveals about one individual.
• Secure aggregation and multiparty computation Combining model updates or statistics so no party sees another party's inputs.
• De-identification Removing or transforming identifiers with documented re-identification risk.
• ML in future networks framework A reference for where ML pipelines sit in network architectures.
• Synthetic data Generated data for training and testing, with documented fidelity and privacy limits.