DeFiRank: federated pairwise ranking for governance in decentralized finance
The paper introduces DeFiRank, a federated learning–based pairwise ranking framework that lets multiple institutions collaboratively train a provider-ranking model for DeFi governance without sharing private transaction data. Evaluated on a real-world blockchain transaction dataset, the federated approach attains NDCG@15 ≈ 0.87 (vs. centralized ≈ 0.92) and Kendall τ ≈ 0.70, and yields governance actions including 648 provider promotions and 39 demotions while preserving data privacy.