In recent years, opportunistic traders have extracted hundreds of millions of dollars from blockchains by reordering financial transactions. The problem stems from the fact that blockchains implement a state machine replication that orders transactions in any consistent order, regardless of the order in which these transactions were received. Existing attempts at enforcing the order perceived by honest participants suffer from cyclic dependencies or message delays. In this paper, we propose the Asynchronous Ordered Atomic Broadcast (AOAB) protocol. It does not suffer from cyclic dependencies or message delays because (i) it assigns an absolute timestamp to transactions, and (ii) it tolerates unbounded message delays. Besides being the first protocol to solve this problem, AOAB is communication-optimal and resilience-optimal. In particular, AOAB makes use of threshold signatures and information dissemination to reach a communication complexity of$\mathcal{O}(n\ell+\lambda n^{2})$, where$n$is the number of processes,$\ell$is the input (transaction) size and$\lambda$is the security parameter. This is optimal when$\ell\geq\lambda n$,

Reordering blockchain transactions to manipulate markets profited hackers by hundreds of millions of dollars. Because they rely on State Machine Replication (SMR), blockchains order transactions without preventing hackers from influencing the chosen order. Some order-fair consensus protocols, like Pompē [33], order transactions before agreeing on this order. They are insufficient because a hacker can leverage the lack of triangle inequality among network latencies to observe pending transactions before issuing their own. Other DAG-based protocols, like Fino [24], use commit-reveal to obfuscate transactions, but cannot prevent reordering by a Byzantine leader.In this paper, we present Lyra, a protocol that solves this problem. The key idea is the combination of a commit-reveal protocol to obfuscate transaction payloads, and a leaderless ordered consensus protocol that predicts the order of transactions. Lyra has optimal good-case latency, prevents reordering attacks, and is scalable. Finally, it outperforms the latency of Pompē by up to 2 times and its throughput by up to 7 times on a 100-node network over 3 continents.

The problem of transaction reordering in blockchains, also known as the blockchain anomaly [Christopher Natoli and Vincent Gramoli, 2016], can lead to fairness limitations [Kelkar et al., 2020] and front-running activities [Philip Daian et al., 2020] in cryptocurrency. To cope with this problem despite f < n/3 byzantine processes, Zhang et al. [Zhang et al., 2020] have introduced the ordering linearizability property ensuring that if two transactions or commands are perceived by all correct processes in the same order, then they are executed in this order. They proposed a generic distributed protocol that first orders commands and then runs a leader-based consensus protocol to agree on these orders, hence requiring at least 11 message delays. In this paper, we parallelize the ordering with the execution of the consensus to require only 6 message delays. For the ordering, we introduce the ordered reliable broadcast primitive suitable for broadcast-based cryptocurrencies (e.g., [Daniel Collins et al., 2020]). For the agreement, we build upon the DBFT leaderless consensus protocol [Tyler Crain et al., 2018] that was recently formally verified [Bertrand et al., 2021]. The combination is thus suitable to ensure ordering linearizability in consensus-based cryptocurrencies (e.g., [Tyler Crain et al., 2021]).