Mecon: A GNN-based graph classification framework for MEV activity detection

The emergence of Maximum Extractable Value (MEV) within blockchain ecosystems poses a formidable challenge, undermining the foundational principles of system integrity, including fairness, security, and user experience. The inherent technological complexity of MEV, compounded by the dynamic nature of blockchain transactions, introduces significant challenges that remain insufficiently addressed in current scholarly discourse. While previous research has explored the use of graph neural networks (GNNs) for MEV detection, these methods often struggle with scalability and adapting to the ever-changing transaction patterns. To address these challenges, we propose a novel framework, Mecon, designed to detect and identify MEV activities by transforming raw blockchain transactional data into an intricate user behavior graph model. This transformation allows for the extraction of complex patterns associated with MEV activities. Additionally, our approach leverages advanced graph-convolutional techniques, significantly enhancing the efficiency and effectiveness of MEV behavior feature extraction. Our empirical analysis reveals that Mecon markedly outperforms existing detection methods in accurate and scalable identification of MEV activities. The framework achieves an impressive F1 score of 94.33%, showing a significant breakthrough in the field of MEV behavior identification and analysis. This research not only provides a robust tool for detecting and analyzing complex transactional behaviors in decentralized networks, but also underscores the potential of deep learning in enhancing blockchain security and fairness. •Introduce Mecon, the first GNN and account-based approach for MEV detection.•Derive features from raw blockchain data to facilitate anomalous behavior detection.•Transform blockchain data centered on addresses into graph structures for analysis.•Propose a two-layer GCN framework to process intricate relational data in blockchain.•Detect MEV behaviors in Ethereum with a remarkable F1 score of 94.33% using Mecon.