Spectral Sentinel: Scalable Byzantine-Robust Decentralized Federated Learning via Sketched Random Matrix Theory on Blockchain

2025/12/14 by Mishra, Animesh
#Distributed #FOS: Computer and information sciences #Machine Learning (cs.LG) #Parallel #and Cluster Computing (cs.DC)

paper · doi:10.48550/arxiv.2512.12617

Abstract

Decentralized federated learning (DFL) enables collaborative model training without centralized trust, but it remains vulnerable to Byzantine clients that poison gradients under heterogeneous (Non-IID) data. Existing defenses face a scalability trilemma: distance-based filtering (e.g., Krum) can reject legitimate Non-IID updates, geometric-median methods incur prohibitive O(n² d) cost, and many certified defenses are evaluated only on models below 100M parameters. We propose Spectral Sentinel, a Byzantine detection and aggregation framework that leverages a random-matrix-theoretic signature: honest Non-IID gradients produce covariance eigenspectra whose bulk follows the Marchenko-Pastur law, while Byzantine perturbations induce detectable tail anomalies. Our algorithm combines Frequent Directions sketching with data-dependent MP tracking, enabling detection on models up to 1.5B parameters using O(k²) memory with k ≪ d. Under a (σ,f) threat model with coordinate-wise honest variance bounded by σ² and f < 1/2 adversaries, we prove (ε,δ)-Byzantine resilience with convergence rate O(σf / √(T) + f² / T), and we provide a matching information-theoretic lower bound Ω(σf / √(T)), establishing minimax optimality. We implement the full system with blockchain integration on Polygon networks and validate it across 144 attack-aggregator configurations, achieving 78.4 percent average accuracy versus 48-63 percent for baseline methods.

Citations

• FLTrust: Byzantine-robust Federated Learning via Trust Bootstrapping
• Advances and Open Problems in Federated Learning
• Local Model Poisoning Attacks to Byzantine-Robust Federated Learning
• Robust and Communication-Efficient Federated Learning from Non-IID Data
• LEAF: A Benchmark for Federated Settings
• Analyzing Federated Learning through an Adversarial Lens
• Mitigating Sybils in Federated Learning Poisoning
• How To Backdoor Federated Learning
• Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates
• The Hidden Vulnerability of Distributed Learning in Byzantium
• Differentially Private Federated Learning: A Client Level Perspective
• Resurrecting the sigmoid in deep learning through dynamical isometry: theory and practice
• Gradient Sparsification for Communication-Efficient Distributed Optimization
• Simple and Deterministic Matrix Sketching
• Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions
• Phase transition of the largest eigenvalue for non-null complex sample covariance matrices
• The Byzantine Generals Problem

Related

• FedCut: A Spectral Analysis Framework for Reliable Detection of Byzantine Colluders
• SpectralKrum: A Spectral-Geometric Defense Against Byzantine Attacks in Federated Learning
• BRFL: A Blockchain-based Byzantine-Robust Federated Learning Model
• Byzantine-resilient Decentralized Stochastic Gradient Descent
• FedGuard: A Diverse-Byzantine-Robust Mechanism for Federated Learning with Major Malicious Clients