Federated Distributed Key Generation

2025-02-28

AI Summary

Traditional distributed key generation (DKG) assumes a static participant set and global threshold, rendering it ill-suited for dynamic, open networks with frequent membership changes and heterogeneous trust assumptions. To address this, we propose Federated DKG (FDKG), wherein each participant autonomously selects its guardian set and defines a local reconstruction threshold—enabling decentralized, fine-grained key distribution and reconstruction. FDKG is the first DKG framework to integrate federated Byzantine agreement principles; it formally defines and guarantees liveness, privacy, and correctness under dynamic settings. Its lightweight, two-round protocol combines secret sharing, polynomial interpolation, and Groth16 zero-knowledge proofs. Experiments show that with 100 participants, 50% participation rate, and 40 guardians per participant, communication complexity is O(nk), with distribution and reconstruction overheads of 332.7 kB and 416.6 kB, respectively, while Groth16 proof generation takes only 0.619 s—significantly enhancing robustness in unreliable environments.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Search and Optimization: Distributed Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Decentralized Web and Fediverse systems
• Security and Privacy: Blockchains and distributed ledgers
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Distributed Key Generation (DKG) is vital to threshold-based cryptographic protocols such as threshold signatures, secure multiparty computation, and i-voting. Yet, standard $(n,t)$-DKG requires a known set of $n$ participants and a fixed threshold $t$, making it impractical for public or decentralized settings where membership and availability can change. We introduce Federated Distributed Key Generation (FDKG), which relaxes these constraints by allowing each participant to select its own guardian set, with a local threshold to reconstruct that participant's partial key. FDKG generalizes DKG and draws inspiration from Federated Byzantine Agreement, enabling dynamic trust delegation with minimal message complexity (two rounds). The protocol's liveness can tolerate adversary that controls up to $k - t + 1$ nodes in every guardian set. The paper presents a detailed protocol, a formal description of liveness, privacy, and integrity properties, and a simulation-based evaluation showcasing the efficacy of FDKG in mitigating node unreliability. In a setting of 100 parties, a 50% participation rate, 80% retention, and 40 guardians, the distribution phase incurred a total message size of 332.7 kB ($O(n,k)$), and reconstruction phase 416.56 kB ($O(n,k)$. Groth16 client-side proving took about 5 s in the distribution phase and ranged from 0.619 s up to 29.619 s in the reconstruction phase. Our work advances distributed cryptography by enabling flexible trust models for dynamic networks, with applications ranging from ad-hoc collaboration to blockchain governance.

Problem

• Enables dynamic trust delegation in decentralized networks
• Reduces message complexity in distributed key generation
• Supports flexible membership and availability in cryptographic protocols

Innovation

• Federated Distributed Key Generation (FDKG) introduced
• Dynamic trust delegation with minimal message complexity
• Simulation-based evaluation shows FDKG mitigates node unreliability