Beyond Noise: Privacy-Preserving Decentralized Learning with Virtual Nodes

arXiv:2404.09536v1 Announce Type: cross
Abstract: Decentralized learning (DL) enables collaborative learning without a server and without training data leaving the users' devices. However, the models shared in DL can still be used to infer training data. Conventional privacy defenses such as differential privacy and secure aggregation fall short in effectively safeguarding user privacy in DL. We introduce Shatter, a novel DL approach in which nodes create virtual nodes (VNs) to disseminate chunks of their full model on their behalf. This …

aggregation arxiv beyond can cs.ai cs.cr cs.dc cs.lg data decentralized defenses devices differential privacy effectively nodes noise privacy server training training data virtual