In today’s fast-paced financial landscape, real-time fraud detection has become essential for safeguarding customer assets and preserving trust in financial institutions. One leading investment bank recently embraced an innovative solution to improve its fraud detection capabilities significantly. By adopting TigerGraph, a powerful graph database and analytics platform, the bank effectively anonymized and analyzed vast amounts of data, elevating its fraud detection system to new heights.
Business Use Case: A prominent financial institution’s credit card business boasts the advantage of real-time approvals, a feature that sets it apart from competitors. While real-time approvals offer a competitive edge, they also expose the bank to potential card application fraud. Recognizing the importance of strengthening fraud detection capabilities, the bank’s data science team sought a scalable real-time fraud detection engine, accompanied by a robust case management front-end. The system needed to process API calls swiftly, with response times of under 0.5 seconds, while effectively categorizing credit card applications in real-time.
Technical Details: To address the challenge of fraud detection, the financial institution managed a database containing over 100 million credit card applications, some of which were flagged as suspicious or fraudulent. The primary hurdle was to enrich this application data by integrating information from multiple internal and external systems, each with distinct data structures and identifiers. Traditional databases and table joins struggled with this complexity, but TigerGraph proved exceptionally adept at handling heterogeneous schemas, making data integration seamless.
TigerGraph’s unique approach involved rigorous testing of each new credit card application against the existing dataset to identify patterns indicative of suspicious activity. As new applications were submitted, TigerGraph dynamically created connections and links between the application data and the existing dataset, based on shared attributes like names or device IDs. These connections formed clusters of related cards, and if any of these clusters contained known suspicious cards, the new application was flagged as potentially fraudulent.
To meet the demanding real-time processing requirements, the financial institution customized TigerGraph’s graph algorithms, adopting a real-time version of the Weakly Connected Components algorithm. This tailored algorithm was specifically designed to deliver rapid results, efficiently analyzing data for potential fraud. By leveraging the open-source and fully functional programming language, GSQL, provided by TigerGraph, the institution adapted the algorithm seamlessly for this unique use case, ensuring that the system could render a fraud decision within the customer’s time constraints.
Additionally, the system supported more comprehensive investigations of flagged applications. The subgraph generated by TigerGraph offered crucial insights to the fraud investigation team, including the distance to suspected fraudulent applications, the number of shared attributes, and the quantity of distinct nodes involved. Armed with this data, investigators conducted further machine learning diagnosis to gain deeper insights and perform in-depth analyses of potential card declines.
Implementation: The financial institution deployed TigerGraph as a distributed cluster on its premises to ensure data privacy and control. This configuration allowed for high availability and performance while mitigating the risk of data loss or service disruptions. The distributed cluster consisted of multiple machines, each equipped with an economical configuration of 8 CPUs and 64GB of RAM. Through two-way replication (4 partitions X 2 replicas), data redundancy and fault tolerance were achieved, guaranteeing the reliability of the system.
By embracing TigerGraph’s cutting-edge graph database and analytics platform, the financial institution made significant strides in strengthening its fraud detection capabilities. The ability to anonymize and analyze vast amounts of data in real-time enabled the institution to protect its customers from potential fraud and solidify its standing as a reliable financial services provider. As the financial industry faces evolving cyber threats, innovative solutions like TigerGraph will continue to play a pivotal role in helping financial institutions stay one step ahead of fraudsters, safeguard customer interests, and maintain trust in the digital age.
Want to see similar results? Contact our team today at info@localhost.
