A PUF-based Secure Aggregation Protocol for Deep Edge Devices

Secure Aggregation (SA) is a key enabler for privacy-preserving data consolidation and Federated Learning in deep edge deployments, where numerous heterogeneous devices combine locally computed values under strict computational and bandwidth constraints. Existing protocols, however, require extensive pairwise key establishment and strong reliance on Public Key Infrastructures or centralized servers, limiting scalability and increasing provisioning complexity. To address these limitations, we introduce a lightweight SA protocol based solely on XOR and addition operations, leveraging Physical Unclonable Functions (PUFs) to derive device-specific secrets without storing long-term keys. Formal analysis shows that, with respect to the number of aggregating nodes J, the protocol achieves network and computational overheads of O(J) and O(1), compared to O(J2) and O(J) in prior approaches. Experimental validation on microcontrollers and ns-3 simulations confirms linear scalability and millisecond-level execution, demonstrating suitability for resource-constrained edge environments.