The rapid evolution of financial technology (FinTech) has brought unprecedented convenience but also given rise to sophisticated and often concealed financial fraud activities. Traditional anti-fraud models

relying on static rules or shallow statistical methods

are increasingly inadequate in detecting complex

syndicated

and cross-platform fraud within vast

high-dimensional

and relationally intricate transaction data. This paper proposes a novel intelligent financial anti-fraud framework that deeply integrates Blockchain technology with Graph Neural Networks (GNNs). Blockchain ensures the immutability

transparency

and traceability of financial transaction data

establishing a trusted data foundation. Subsequently

GNNs leverage their powerful capabilities in analyzing graph-structured data to uncover intricate relational patterns and hidden fraud communities that are often missed by conventional approaches. We detail a methodology for constructing dynamic graphs from blockchain-recorded time-series data and designing a GNN model capable of identifying evolving fraudulent activities. Through conceptual validation and simulated experiments

this research aims to demonstrate how this integrated approach can significantly enhance the accuracy

robustness

and interpretability of financial anti-fraud systems. The proposed framework holds substantial academic value by bridging existing research gaps in hybrid Blockchain-AI applications for FinTech and offers practical implications for bolstering financial market stability and security.