Byzantine-Resilient Federated Learning via QUBO-Based Client Selection on Quantum Annealers

In three linesQuantum annealing approach for selecting trustworthy clients in federated learning against Byzantine attacks. Reformulates client selection as QUBO problem jointly optimizing closest client subsets. MultiSignal ensemble achieves 95.3% detection accuracy at 100 clients on MNIST vs 91.8% for classical MultiKrum.

Summary generated by Claude — human-verified