FlagThis

Abstract

Blockchain consensus faces a trilemma of security, latency, and decentralization. High-throughput systems often require a reduction in decentralization or robustness against strong adversaries, while highly decentralized and secure systems tend to have lower performance. We present BlueBottle, a two-layer consensus architecture to tackle these challenges. The main finalization layer, which we call the core layer, is run in a medium-sized validator set. It is instantiated with our novel BB-Core consensus protocol. BB-Core exposes two client finality paths following the FlexibleBFT [44] paradigm. Fast-path clients that assume __ = 5 __ + 1 experience 2 __ -finality latency. Our experiments show that BB-Core reduces latency by 2025% compared to Mysticeti. Checkpoint-path clients wait for a checkpoint certificate (4 __ latency), which tolerates an additional _ <_ 2 __ alive-but-corrupt participantsachieving a total of 60% equivocation-fault safety. The key mechanism is a single-proposal-per-height rule that guarantees at most one valid checkpoint exists per slot height, even under heavy equivocation. The secondary finalization layer, BB-Guard, is run along a highly decentralized set of guard validators (or full nodes) and provides synchronous decentralized timestamping, proactive misbehavior detection, and a synchronous recovery path. When guard validators observe equivocations or liveness failures they disseminate evidence, agree on misbehaving parties for exclusion or slashing, and either restart the core protocol (for liveness violations) or select a canonical fork (for safety violations). Clients can choose to accept the synchrony assumption as their main assumption and withstand a __ = 2 __ + 1 security level. BB-Guard can be deployed along any consensus protocol. In BlueBottle combining BB-Core with BB-Guard allows clients to choose their security level against their latency requirements.