Paper 2024/1428

Mario: Multi-round Multiple-Aggregator Secure Aggregation with Robustness against Malicious Actors

Abstract

Federated Learning (FL) enables multiple clients to collaboratively train a machine learning model while keeping their data private, eliminating the need for data sharing. Two common approaches to secure aggregation (SA) in FL are the single-aggregator and multiple-aggregator models. Existing multiple-aggregator protocols such as Prio (NSDI 2017), Prio+ (SCN 2022), Elsa (S\&P 2023) either offer robustness only in the presence of semi-honest servers or provide security without robustness and are limited to two aggregators. We introduce Mario, the first multi-aggregator SA protocol that is both secure in a malicious setting and provides robustness. Similar to prior work of Prio and Prio+, Mario provides secure aggregation in a setup of $n$ servers and $m$ clients. Unlike previous work, Mario removes the assumption of semi-honest servers, and provides a complete protocol with robustness against less than $n/2$ malicious servers, defense with input validation of upto $m-2$ corrupted clients, and dropout of any number of clients. Our implementation shows that Mario is $3.40\times$ and $283.4\times$ faster than Elsa and Prio+, respecitively.