Paper 2026/1570

Anchor-DKG: Distributed Key Generation with Repeating Parties

Hanwen Feng, University of Canterbury
Qiang Tang, University of Sydney
Sri AravindaKrishnan Thyagarajan, University of Sydney
Abstract

A party may participate in multiple threshold cryptosystems. For example, it may serve on multiple overlapping threshold committees in a proof-of-stake blockchain or a distributed oracle network, or act as a client of multiple cryptocurrency wallet services built on threshold cryptography. With conventional distributed key generation (DKG), each threshold system independently generates its key shares, imposing significant key-management overhead on such a repeating party. In contrast, modern key-management practice favors deriving all cryptographic material deterministically from a single master key, raising a fundamental question: Can DKG be reconciled with key derivation while preserving security and compatibility with legacy threshold systems? We present Anchor-DKG, a new DKG protocol that allows up to $t^{\mathsf{rec}}$ (the reconstruction threshold) parties to deterministically fix their secret key shares while retaining standard security guarantees. Anchor-DKG supports concurrent executions with overlapping participants across multiple DKG instances and remains fully compatible with legacy threshold schemes, including ECDSA, BLS, Schnorr, and ElGamal. At the core of Anchor DKG lies a new technique: fixed-point distributed polynomial sampling (FpDpS). FpDpS allows parties to jointly sample a random $(t^{\mathsf{rec}}-1)$-degree polynomial $f$ such that $f(i) = s_i$ at designated points $i$, where each $s_i$ can be a private input, e.g., a key derived from a master secret. The final secret key remains $f(0)$, ensuring compatibility with existing discrete-log-based threshold systems. We provide an efficient construction of Anchor DKG under standard cryptographic assumptions, which, compared to classical constructions such as Gennaro et al. (J.Cryptol. 2007), only incurs one more point-to-point round and marginal computation. Experimental results show that, for a network size of $n=128$, our protocol incurs a per-party computation cost of $1.59$ s, compared to $1.36$ s for GJKR.

Metadata
Available format(s)
PDF
Category
Cryptographic protocols
Publication info
Published elsewhere. Major revision. ACM CCS 2026
Keywords
Distributed Key GenerationThreshold CryptographyKey DerivationKey Management
Contact author(s)
hanwen feng @ canterbury ac nz
qiang tang @ sydney edu au
aravind thyagarajan @ sydney edu au
History
2026-08-03: revised
2026-07-31: received
See all versions
Short URL
https://ia.cr/2026/1570
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2026/1570,
      author = {Hanwen Feng and Qiang Tang and Sri AravindaKrishnan Thyagarajan},
      title = {Anchor-{DKG}: Distributed Key Generation with Repeating Parties},
      howpublished = {Cryptology {ePrint} Archive, Paper 2026/1570},
      year = {2026},
      url = {https://eprint.iacr.org/2026/1570}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.