A Fraud Detection System in Financial Networks Using AntiBenford Subgraphs and Machine Learning Algorithms

Detecting financial fraud within banking systems and digital transactional ledgers is a complex challenge that demands robust methodologies. One such approach is graph mining, which delves into datasets to reveal patterns and relationships, particularly in financial networks. The AntiBenford subgraph, inspired by Benford's Law, is tailored for identifying irregularities in financial transactions. However, it has been known to produce false positives. To address this issue, a proposed solution combines the AntiBenford subgraph with an unsupervised Machine Learning Algorithm (MLA). This innovative Fraud Detection System significantly bolsters security within financial networks. Scrutinizing leading digits and examining deviations from Benford's Law expectations effectively pinpoints anomalies in numerical data, a crucial step in identifying potential fraudulent activities. Thanks to its incorporation of MLA techniques, the system adapts and refines its capabilities over time, making it a powerful tool for early detection of financial misconduct. This research contributes a valuable asset to the financial sector, reinforcing security and trust by enabling the early detection of fraudulent transactions. By leveraging Benford's Law and machine learning, this system plays a pivotal role in safeguarding against financial misconduct, ultimately providing a 94.83% accuracy rate in fraud detection, a testament to its effectiveness.