AI RESEARCH
Beyond Passive Aggregation: Active Auditing and Topology-Aware Defense in Decentralized Federated Learning
arXiv CS.LG
•
ArXi:2603.18538v1 Announce Type: new Decentralized Federated Learning (DFL) remains highly vulnerable to adaptive backdoor attacks designed to bypass traditional passive defense metrics. To address this limitation, we shift the defensive paradigm toward a novel active, interventional auditing framework. First, we establish a dynamical model to characterize the spatiotemporal diffusion of adversarial updates across complex graph topologies. Second, we