Paper 2026/892

Adaptive Distributed Key Generation for Discrete-Log Cryptosystems

Ruben Baecker, Friedrich-Alexander-Universität Erlangen-Nürnberg
Paul Gerhart, TU Wien
Stanislaw Jarecki, University of California, Irvine
Phillip Nazarian, University of California, Irvine
Daniel Rausch, University of Stuttgart
Dominique Schröder, TU Wien
Abstract

Threshold signatures are widely deployed in decentralized asset custody and blockchain infrastructure to eliminate single points of failure. In these systems, an attacker can adaptively corrupt committee members based on public membership and protocol transcripts. Recent cryptanalysis shows this threat is concrete: schemes exposing per-party key commitments enable practical forgeries for committees of roughly 200 signers at the cost of just one minute of the global Bitcoin hashrate. While a new generation of signing schemes defeats this attack by hiding individual key shares, every distributed key generation (DKG) protocol used in practice exposes this information, voiding the adaptive security of any scheme built on top. Trusted dealers avoid the leakage but reintroduce the single point of failure that threshold cryptography exists to eliminate. True adaptive security must therefore span both key generation and signing. We present Janus, the first DKG suite to achieve this end-to-end guarantee. Janus serves as a drop-in replacement for a trusted dealer in threshold schemes, hiding all key shares while outputting a standard DLog public key compatible with existing verifiers, wallets, and smart contracts. Assuming secure erasures, we prove security against an adaptive adversary that corrupts members mid-protocol and controls a dishonest majority. We also show that Janus directly supports a variety of DLog-based primitives, including threshold Schnorr signatures and oblivious pseudorandom functions. Our suite includes Janus2R, which achieves an unbiased key in two rounds, and Janus1R, which completes within a single round by tolerating an additive key shift. When a node misbehaves, both variants generate a 259-byte universally verifiable proof for automated on-chain slashing, and both naturally allow committees to proactively refresh their shares without changing the public key. Our open-source Rust implementation completes key generation for a 16-node intercontinental committee in 310 ms and scales to 512 nodes.

Metadata
Available format(s)
PDF
Category
Cryptographic protocols
Publication info
Preprint.
Keywords
Distributed Key GenerationAdaptive SecurityUniversal Composability
Contact author(s)
ruben baecker @ fau de
paul gerhart @ tuwien ac at
sjarecki @ uci edu
pnazaria @ uci edu
daniel rausch @ sec uni-stuttgart de
dominique schroeder @ tuwien ac at
History
2026-10-06: revised
2026-05-06: received
See all versions
Short URL
https://ia.cr/2026/892
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2026/892,
      author = {Ruben Baecker and Paul Gerhart and Stanislaw Jarecki and Phillip Nazarian and Daniel Rausch and Dominique Schröder},
      title = {Adaptive Distributed Key Generation for Discrete-Log Cryptosystems},
      howpublished = {Cryptology {ePrint} Archive, Paper 2026/892},
      year = {2026},
      url = {https://eprint.iacr.org/2026/892}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.