Publication

Abstract : This paper presents a new way to spot fraudulent transactions using Graph Neural Networks (GNNs). The transactions are treated like a network, where people or accounts are connected by their transactions. GNNs help to analyze the network to find suspicious patterns that might indicate fraud. By looking at both local and broader connections, the proposed method catches unusual activities effectively. The proposed approach was tested on a large set of real transaction data, and it was found to be better than traditional methods, with higher accuracy and fewer mistakes. The findings show that GNNs can be a powerful tool for building stronger, adaptable fraud detection systems to protect financial systems