DPLE: a Privacy-Enhanced and Straggler-Resilient Distributed Learning Framework for Smart Cloud

EasyChair Preprint no. 12740, version 2

Abstract

In the intelligent cloud setting, distributed learning encounters privacy and straggler challenges. Lagrange coded computing offers partial relief. Yet, if the number of inquisitive but honest nodes surpasses a threshold or if there are external eavesdroppers, system privacy becomes compromised. To tackle this issue, we introduce a novel approach called DPLE (Differentially Private Lagrange Encoding). Additionally, we provide theoretical analyses to determine the artificial noise variance necessary for achieving desired privacy levels within this framework. Through experimentation, we demonstrate the efficacy of our approach and evaluate how different system parameters affect accuracy.

Keyphrases: artificial noise, differential privacy, distributed learning, Lagrange coded computing

@Booklet{EasyChair:12740,
  author = {Yilei Xue and Jianhua Li and Jun Wu},
  title = {DPLE: a Privacy-Enhanced and Straggler-Resilient Distributed Learning Framework for Smart Cloud},
  howpublished = {EasyChair Preprint no. 12740},

  year = {EasyChair, 2024}}