FedSharp: Byzantine Robust Decentralized Federated Learning Under Noisy Channel

Decentralized federated learning (DFL) eliminates the central aggregation server but remains vulnerable to data heterogeneity, Byzantine attacks, and channel noise. We show that noise accumulates as models propagate across nodes, becoming increasingly significant as the learning rate decays.

去中心化联邦学习（DFL）消除了中央聚合服务器，但仍然容易受到数据异构性、拜占庭攻击和信道噪声的影响。我们证明了随着模型在节点间传播，噪声会不断累积，并且随着学习率的衰减，这种影响变得愈发显著。

Moreover, accumulated noise enables Byzantine attackers to exploit loss-based model selection by mimicking honest model averages. To address these challenges, we propose screened, hop-aware, ranked, proximal averaging, a robust DFL framework (FedSharp).

此外，累积的噪声使拜占庭攻击者能够通过模仿诚实模型的平均值来利用基于损失的模型选择。为了应对这些挑战，我们提出了一种名为“筛选、跳数感知、排序、近端平均”的鲁棒 DFL 框架（FedSharp）。

In this approach, each node maintains models from its 2b predecessors, ensuring at least b honest candidates when at most b nodes are Byzantine. FedSharp screens suspicious models using coordinate-wise median deviations, averages the b lowest-loss candidates to reduce fresh channel-noise variance by a factor of b, and applies noise-adaptive gradient-norm regularization and proximal training to mitigate accumulated noise and non-IID drift.

在这种方法中，每个节点维护来自其 2b 个前驱节点的模型，从而确保在最多有 b 个节点为拜占庭节点时，至少有 b 个诚实候选模型。FedSharp 使用坐标中位数偏差来筛选可疑模型，对 b 个损失最低的候选模型进行平均以将新鲜信道噪声方差降低 b 倍，并应用噪声自适应梯度范数正则化和近端训练来减轻累积噪声和非独立同分布（non-IID）漂移。

Experiments on a 100-node ring with 33 Byzantine nodes show that FedSharp achieves 93.5-94.9% accuracy on MNIST, compared with 29.4-38.1% for a state-of-the-art single-model-selection benchmark, and 32.8-48.8% on CIFAR-10, compared with 10.0-30.5%, across Gaussian and hidden attacks under IID and non-IID data. On MNIST, FedSharp also achieves accuracy within approximately one percentage point of the benchmark operating without channel noise.

在包含 33 个拜占庭节点的 100 节点环形网络上的实验表明，FedSharp 在 MNIST 数据集上达到了 93.5-94.9% 的准确率，而最先进的单模型选择基准仅为 29.4-38.1%；在 CIFAR-10 数据集上，FedSharp 达到了 32.8-48.8% 的准确率，而对比基准为 10.0-30.5%。这些结果涵盖了在 IID 和非 IID 数据下针对高斯攻击和隐藏攻击的测试。在 MNIST 上，FedSharp 的准确率也达到了在无信道噪声环境下运行的基准测试的约一个百分点以内。