Federated Multi-Armed Bandits Under Byzantine Attacks

Understand

Multi-armed bandits (MAB) is a sequential decision-making model in which the learner controls the trade-off between exploration and exploitation to maximize its cumulative reward.

• Federated multi-armed bandits (FMAB) is an emerging framework where a cohort of learners with heterogeneous local models play an MAB game and communicate their aggregated feedback to a server to learn a globally optimal arm.
• Two key hurdles in FMAB are communication-efficient learning and resilience to adversarial attacks.
• To address these issues, we study the FMAB problem in the presence of Byzantine clients who can send false model updates threatening the learning process.