International Association
for Cryptologic Research

Dynamic participation has recently become a key requirement to devise permissionless consensus protocols, as it adds a degree of robustness to events that include portions of participants going offline, preserving safety and liveness of such dynamically available protocols. This notion, formalized by Pass and Shi (ASIACRYPT 2017) with the sleepy model, has been implicitly adopted to model several blockchain protocols such as, for example, the Ethereum's consensus protocol, Gasper.

Neu, Tas, and Tse (S&P 2021) show that LMD-GHOST, the dynamic availability component of Gasper, is actually not secure even in a context of full-participation, i.e., with all the validators online. Mitigations have shortly after been developed to cope with its problems, but the resulting protocol still falls short of achieving dynamic availability, motivating the research of more secure dynamically available protocols.

In this work we present RLMD-GHOST, a synchronous dynamically available protocol that does not lose safety during bounded periods of asynchrony. This protocol results appealing especially for practical systems, where strict synchrony assumptions might not always hold, contrary to what is generally assumed with standard synchronous protocols. Moreover, we introduce the generalized sleepy model, in which our results will be proved. This model takes up from the original sleepy model presented by Pass and Shi and extends it with more generalized and stronger constraints in the corruption and sleepiness power of the adversary. This allows us to explore a broad space of dynamic participation regimes which falls between complete dynamic participation and no dynamic participation, i.e., with every participant online, offering a foundation for the analysis of dynamic available protocols.