How an Online Bank Exposes Complex Fraud with Graph Database Technology

The Challenge

BforBank faced a real challenge identifying complex fraud schemes in siloed data. Every day, BforBank’s database management systems collect large amounts of data. From transactions and money flows to Know Your Customer (KYC) documents, the bank has large volumes of structured and unstructured data distributed across multiple silos. Monitoring this data is critical to lowering risks and financial losses.

To investigate flagged customers, transactions or behaviors, the bank’s risk and compliance team was using a bank fraud solution built on relational technology. As a result, querying connections within the data to confirm fraudulent activities or uncover fraud rings was a tedious, long and sometimes unsuccessful process.

“A request could take from two minutes to several hours when querying for connections across multiple relational tables. Cross-field queries could take several days,” said Alexandre Dressayre, Compliance Officer at BforBank.

Complex cases required access to information scattered across data silos. Some types of fraud, such as phishing, required the intervention of the IT Department.

Every time investigators had to request additional technical resources, this slowed down the fraud investigation process, potentially resulting in larger losses. The bank needed a more effective way of monitoring all its data to reduce risk and financial losses.

The Solution

To improve fraud detection and reduce investigation time, BforBank looked to graph database technology. Linkurious provided a bundled solution with Linkurious Enterprise software, which offered off-the-shelf visualization and analysis on top of Neo4j’s graph database. According to Dressayre, it was the perfect fit.

BforBank’s risk and compliance team started by designing a data model, loading all customer data, bank transfer orders, check cashing activities and IP addresses into Neo4j. The graph data was instantly available, providing an intuitive interface to investigate the hidden connections of suspicious clients.

“Thanks to the available network of data, we can spread out the connections and try to find if a fraudster is connected to other clients, through IP, postal or email addresses, for example. This helps us detect fraud rings or identity theft fraud,” said Dressayre.

Now the BforBank team is able to detect fraud patterns that were too complex to identify in the past.

“The first pattern we set up was one related to phishing fraud. The system reports cases where clients have multiple and suspicious connection behaviors,” said Dressayre.

And, as new fraud schemes are identified, BforBank can set up additional alerts to oppose those threats.

Download Case Study