Lotto: Secure Participant Selection against Adversarial Servers in Federated Learning

Watch a 12-minute conference presentation from USENIX Security '24 exploring a novel Federated Learning (FL) system called Lotto, designed to ensure secure participant selection when dealing with adversarial servers. Learn how researchers from Hong Kong University of Science and Technology, University of Michigan, and Nokia Bell Labs address the critical challenge of maintaining security guarantees in FL systems where servers cannot be trusted. Discover how Lotto implements both random and informed selection algorithms, enabling clients to autonomously determine their participation using verifiable randomness while preventing adversarial servers from creating a dishonest majority through strategic client selection. Examine theoretical analysis demonstrating how Lotto effectively maintains the proportion of compromised participants at the natural base rate, along with experimental results showing comparable time-to-accuracy performance to traditional insecure selection methods.