Graph Neural Networks (GNNs) have garnered significant attention in recent years for their advanced data processing capabilities, achieving remarkable results in tasks like node classification, link prediction, and graph classification. These models have proven highly effective in tackling complex problems, particularly in financial applications. Simultaneously, the rise of transaction fraud on e-commerce platforms has underscored the critical need for robust anomaly detection (AD) solutions across various industries. This growing demand has driven researchers to explore GNNs for AD, leveraging ongoing advancements in deep learning. In this paper, we present an entity-based GNN model designed for anomaly detection, focusing on identifying fraudulent users and behaviors. Our approach incorporates both entity relationships and transaction correlations as key information to uncover anomalies. Comprehensive experiments on public datasets demonstrate the effectiveness and superior performance of our proposed model.