网络安全与数据治理2025，Vol.44Issue(7)：20-26,7.DOI:10.19358/j.issn.2097-1788.2025.07.004
基于深度强化学习的以太坊MEV交易防护与交易排序优化
Ethereum MEV transaction protection and transaction ordering optimization based on deep reinforcement learning
严彦胜 1李京1
作者信息
- 1. 中国科学技术大学 计算机科学与技术学院,安徽 合肥 230026
- 折叠
摘要
Abstract
The problem of Maximal Extractable Value(MEV)in Ethereum transaction ordering allows malicious actors to profit by manipulating transaction sequences,undermining network fairness and increasing Gas fees.To suppress MEV behavior and op-timize fairness while enhancing system efficiency,this paper proposes a transaction ordering optimization method based on Deep Q-Network(DQN).By designing appropriate state space,action space,and reward function,the agent can autonomously learn optimal ordering strategies.The effectiveness of the proposed method is systematically validated using a Geth private chain,along with Flashbots MEV-Explore and Ethereum Mempool data.Experimental results show that the DQN-based ordering strategy re-duces the MEV extraction rate to below 13%,decreases average Gas fees by about 33.1%compared to traditional strategies,and raises the fairness index to 0.78,significantly outperforming existing methods.The closed-loop experimental system built in this paper provides a feasible solution for optimizing fairness and efficiency in blockchain transaction ordering.关键词
以太坊/最大可提取价值(MEV)/交易排序优化/深度Q网络(DQN)Key words
Ethereum/Maximal Extractable Value(MEV)/transaction ordering optimization/Deep Q-Network(DQN)分类
信息技术与安全科学引用本文复制引用
严彦胜,李京..基于深度强化学习的以太坊MEV交易防护与交易排序优化[J].网络安全与数据治理,2025,44(7):20-26,7.
