Differential Privacy as a Perk: Federated Learning over Multiple-Access Fading Channels with a Multi-Antenna Base Station
This work investigates how inherent channel noise in over-the-air federated learning (AirFL) can provide differential privacy without adding artificial noise. The study proves that under specific low signal-to-noise ratio conditions, AirFL with differential privacy achieves performance identical to standard AirFL, meaning the cost of privacy is paid for free by the existing channel impairments.