Artificial intelligence and graph analytics are detecting fraudulent transactions more effectively, leveraging the industry's banking sector's capabilities to safeguard their customers, a study by Union Bank of the Philippines showed.
Developed by UnionBank's Artificial Intelligence and Innovation Center of Excellence team, the new method provided nuanced insights into fraud, which can potentially improve systems to mitigate financial risks effectively and enhance decision-making processes, making them more accurate, faster and efficient.
"With this, we have further proven that AI can significantly augment our ability to spot patterns in transaction flows, detect malicious activities, and prioritize suspicious accounts for further investigation," said Dr. Adrienne Heinreich, UnionBank head of AI Center of Excellence, Data and AI Group.
UnionBank used graph analytics to scrutinize the intricate relationships between transactions. For instance, fraudsters often secretly involve intermediary accounts to facilitate money laundering across a network. By expanding their analysis to three degrees of connection instead of just the first degree, the Bank was able to gain a more comprehensive understanding of the risks associated with different fraudulent activities.
"Examining an account's influence within a network is one pattern we seek to identify. The technique we employed to assess this is by measuring its centrality," said UnionBank data scientist Abigail Antenor.
This approach allowed UBP to quantify the connections associated with each account, identify those acting as middlemen, and evaluate their proximity to other accounts to determine the speed of fund transfers. Subsequently, they studied how these measures correlate with fraud in different degrees of connection previously mentioned to determine which indicator is most relevant in those scenarios.
Tailoring fraud indicators to each degree proved effective, as results showed that 19 percent more fraudulent transactions were detected by applying fraud indicators to include second and third degrees, with 80 percent saved for the turnaround time.
This breakthrough is timely, as fraudsters are continuously becoming more innovative in executing attacks, especially those directed at banking systems and their customers.