Trust-Aware Topology Learning for Dynamic Decentralized Federated Learning under Adversaries

2026-08-04

AI Summary

This work addresses the vulnerability of dynamic decentralized federated learning to adversarial nodes that simultaneously manipulate model updates and network topology, thereby undermining collaborative reliability. To counter this threat, the paper proposes the DMTT protocol, which extends MURMURA by introducing a trusted topology mechanism. Specifically, it employs evidential deep learning to assess distributional consistency among peers and integrates link reliability, signed topological claims, and witness verification to construct a trust-aware dynamic graph. Model aggregation is then restricted to a high-trust subgraph. This approach uniquely co-optimizes topological trust modeling and model compatibility, theoretically bounding Byzantine influence within a residual term approaching zero. Experiments on UCI HAR and PAMAP2 datasets demonstrate that even with 80% adversarial nodes, honest participants achieve accuracies exceeding 0.862 and 0.829, respectively—significantly outperforming FedAvg and existing robust aggregation methods—with effectively zero Byzantine weight in practice.

Abstract

In dynamic mobile decentralized federated learning (DFL), adversaries can poison both model updates and the topology information devices use to choose collaborators. We present DMTT (Dynamic MURMURA with Trusted Topology), a decentralized personalized FL protocol built on MURMURA, which uses evidential deep learning to down-weight distribution-mismatched peers, extended here to time-varying graphs under topology-manipulation attacks. Each device maintains a confidence-weighted local topology view from link-reliability estimates, signed topology claims, witness corroboration, and a Beta-distributed source-trust model, then aggregates only over a trust-screened collaborator set using a composite score fusing model compatibility, topology trust, and link reliability. We prove the screened mixing matrices confine Byzantine influence to a bounded residual $δ_{max}$ that vanishes under perfect screening, and implement DMTT as a coordinator-free distributed system with each client running as an independent ZeroMQ process synchronized by a shared wall-clock epoch. On UCI HAR and PAMAP2, each partitioned across 100 mobile clients with Dirichlet heterogeneity, DMTT sustains honest-node accuracy above 0.862 (UCI HAR) and 0.829 (PAMAP2) across all tested adversary fractions (10 to 80%), nearly matching no-attack accuracy at low fractions and degrading gracefully toward local-only performance at extremes; static and dynamic FedAvg collapse to chance at every fraction, and robust aggregators (Krum, BALANCE, UBAR) fail to consistently beat a local-only baseline, while DMTT is the only method that clears this bar across both datasets at all fractions, with surviving Byzantine aggregation weight empirically zero throughout, consistent with $δ_{max}$=0. The protocol runs end-to-end on real nodes via a coordinator-free ZeroMQ backend on the Melbourne Research Cloud.

Problem

• Decentralized Federated Learning
• Adversarial Attacks
• Topology Manipulation
• Trust-aware Learning
• Byzantine Robustness

Innovation

• trust-aware topology
• evidential deep learning
• decentralized federated learning
• Byzantine robustness
• dynamic graph aggregation