Invariant Aggregator for Defending against Federated Backdoor Attacks

arXiv:2210.01834v3 Announce Type: replace
Abstract: Federated learning enables training high-utility models across several clients without directly sharing their private data. As a downside, the federated setting makes the model vulnerable to various adversarial attacks in the presence of malicious clients. Despite the theoretical and empirical success in defending against attacks that aim to degrade models' utility, defense against backdoor attacks that increase model accuracy on backdoor samples exclusively without hurting the utility on other samples remains challenging. To this end, …

abstract adversarial adversarial attacks aim arxiv attacks backdoor cs.cr cs.lg data federated learning private data success training type utility vulnerable