Strategic Peer Selection Using Transaction Value and Latency

Abstract

Many blockchains utilize public peer-to-peer networks to communicate transactions. As activity on blockchain-based DeFi protocols has increased, there has been a sharp rise in strategic behaviour from bots and miners, commonly captured by the notion of Maximal Extractable Value (MEV). While many works have focused on MEV arising from the smart contract layer or consensus layer, in this work we study how a strategic agent can maximise realisable MEV through the optimal choice of network peers. Specifically, we study how existing definitions and algorithms for latency optimization can be augmented with information about the transactions themselves in order to optimize peering algorithms. We formally model this optimization objective for two classes of consensus protocols : 1) time-based ("fair ordering") protocols and 2) single leader-based protocols. We present an efficient local algorithm for choosing peers strategically, and evaluate our algorithm on real world data to show that it outperforms benchmark algorithms that either choose peers randomly or do not exploit information about blockchain transactions.