Cell-free massive mimo boosts privacy in federated learning systems
Privacy-Preserving Over-the-Air Federated Learning over Cell-Free Massive MIMO System with User-Centric Clustering
Information Theory
Summary
Training AI models on your device without sending raw data helps protect your privacy, but making this process very private is tricky. This paper studies a wireless network setup where many small antennas help devices send their learning updates together in a way that naturally hides individual data details using the network’s noise. The researchers find smart ways to decide how devices connect and how much power they use to balance privacy and learning quality. Their experiments show much better accuracy and privacy than older methods on common image tasks.
What this means in practice
- •For wireless network engineers: Design cell-free massive MIMO networks that optimize user connections and transmit powers to enhance privacy during distributed learning without heavy artificial noise addition.
- •For edge ai developers: Implement privacy-preserving federated learning over wireless systems that maintain model accuracy under strict privacy limits using joint optimization of communication parameters.
Authors
Manobendu Sarker, Soumaya Cherkaoui
Abstract
This paper investigates cross-layer design for differentially private federated learning over cell-free massive MIMO networks with over-the-air (OTA) aggregation. Distributed access points (APs) collaboratively serve user equipments (UEs) through user-centric clustering, where clipped local model updates are perturbed with Gaussian noise and simultaneously transmitted over a shared uplink for OTA averaging at a central processing unit. Exploiting channel hardening, we show that thermal noise inherent in wireless reception can act as a privacy-masking resource, reducing or even eliminating the need for artificial noise. Based on this observation, we formulate a joint optimization problem over AP-UE associations, UE transmit powers, and artificial noise variances to minimize aggregation variance subject to per-UE zero-concentrated differential privacy constraints derived from an ergodic privacy accounting framework. Analysis reveals a regime-dependent structure: stringent privacy requirements favor low transmit powers and sparse AP associations to exploit thermal noise for privacy, whereas relaxed requirements favor larger clusters to improve aggregation quality. The resulting mixed-integer nonconvex problem is solved via alternating optimization combining difference-of-convex power allocation and greedy coordinate-descent clustering. Numerical results show up to 89% lower aggregation variance than conventional baselines and enable 97% test accuracy on MNIST with consistent gains on CIFAR-10. Under strict privacy budgets, competing schemes remain near random-chance performance, while under relaxed budgets they continue to lag significantly behind the proposed design.