SuperFL: Privacy-Preserving Federated Learning with Efficiency and Robustness

ePrint Report: SuperFL: Privacy-Preserving Federated Learning with Efficiency and Robustness

Yulin Zhao, Hualin Zhou, Zhiguo Wan

Federated Learning (FL) accomplishes collaborative model training without the need to share local training data. However, existing FL aggregation approaches suffer from inefficiency, privacy vulnerabilities, and neglect of poisoning attacks, severely impacting the overall performance and reliability of model training. In order to address these challenges, we propose SuperFL, an efficient two-server aggregation scheme that is both privacy preserving and secure against poisoning attacks. …

aggregation attacks data efficiency eprint report federated federated learning local model training performance poisoning poisoning attacks privacy report robustness share training training data vulnerabilities wan