Maximal Extractable Value (MEV) activities pose critical operational challenges for blockchain enterprises, requiring automated detection systems to maintain platform integrity and regulatory compliance. Current industrial practices rely on heuristic rule-based methods with substantial accuracy limitations and inability to adapt to evolving MEV strategies. This paper presents an automated software engineering solution for large-scale MEV detection, introducing a novel graph-based profitability identification algorithm that replaces inflexible heuristic rules with adaptive mechanisms. Our automated system achieves 0.6% false positive rates for arbitrage detection and 2.4% false negative rates, significant improvements over existing methods with much higher error rates.

We validate our approach on the most comprehensive dataset to date, i.e., 21 million Ethereum blocks containing 2.5 billion transactions, covering critical infrastructure transitions including The Merge and Proposer-Builder Separation. Our automated pipeline identifies 12.1 million MEV activities, including 1.2 million previously undetectable advanced variants that pose emerging risks to platform operators. Key findings provide actionable insights for blockchain enterprises: private transaction architectures protect 71.4% of low-yield MEV opportunities rather than harming participants, contradicting previous assumptions. However, we identify concerning builder-searcher collusion involving 2,000+ transactions worth 350 ETH, highlighting compliance risks. Additionally, intensifying centralization trends show a single oligopoly controlling 43.1% of MEV activities in 2024, presenting systemic risks.

Our automated detection framework provides blockchain enterprises with production-ready tools for MEV monitoring, risk assessment, and compliance management while offering critical insights for infrastructure design decisions in rapidly evolving DeFi environments.