Skip to main content

Showing 1–1 of 1 results for author: Obioma, C P

.
  1. arXiv:2505.10297  [pdf, other

    cs.LG cs.AI cs.CR

    Defending the Edge: Representative-Attention for Mitigating Backdoor Attacks in Federated Learning

    Authors: Chibueze Peace Obioma, Youcheng Sun, Mustafa A. Mustafa

    Abstract: Federated learning (FL) enhances privacy and reduces communication cost for resource-constrained edge clients by supporting distributed model training at the edge. However, the heterogeneous nature of such devices produces diverse, non-independent, and identically distributed (non-IID) data, making the detection of backdoor attacks more challenging. In this paper, we propose a novel federated repr… ▽ More

    Submitted 15 May, 2025; originally announced May 2025.

    Comments: Submitted to ESORICS 2025