The Impact of Adversarial Node Placement in Decentralized Federated Learning Networks

As Federated Learning (FL) grows in popularity, new decentralized frameworksare becoming widespread. These frameworks leverage the benefits ofdecentralized environments to enable fast and energy-efficient inter-devicecommunication. However, this growing popularity also intensifies the need forrobust security measures. While existing research has explored various aspectsof FL security, the role of adversarial node placement in decentralizednetworks remains largely unexplored. This paper addresses this gap by analyzingthe performance of decentralized FL for various adversarial placementstrategies when adversaries can jointly coordinate their placement within anetwork. We establish two baseline strategies for placing adversarial node:random placement and network centrality-based placement. Building on thisfoundation, we propose a novel attack algorithm that prioritizes adversarialspread over adversarial centrality by maximizing the average network distancebetween adversaries. We show that the new attack algorithm significantlyimpacts key performance metrics such as testing accuracy, outperforming thebaseline frameworks by between 9findings provide valuable insights into the vulnerabilities of decentralized FLsystems, setting the stage for future research aimed at developing more secureand robust decentralized FL frameworks.