In recent years, graph databases have gained significant popularity due to their ability to model and analyse complex, interlinked data structures. This has not only advanced social media analysis but also anti- money laundering (AML). Unlike relational databases which store data in rows and columns, graph databases excel in representing networks of entities and the dynamic relationship between them. This allows graph databases to traverse relationship between entities, and make it possible to discover deeper or hidden relationships and patterns. This is especially beneficial for financial crime compliance, where criminals are using increasingly layered and complexed methods to launder funds. In addition, graph algorithms help to surface patterns and characteristics that are buried in mountains of data, filter through the noise and identify truly suspicious behaviours.
Modelling complex networks of transactions
Money laundering typically involves a complex web of financial transactions, involving multiple intermediaries, accounts and entities. A graph database naturally represents these relationships, amongst other relationships in the form of shared characteristics and attributes. It helps to visualise and make sense of the interconnectedness of the nodes, highlight important nodes and relationships. This can give compliance analysts a more comprehensive view of suspicious activity alerts.
Detection of suspicious patterns and anomalies
Network analysis can easily identify suspicious patterns that are indicative of money laundering, such as:
- Circular transactions: funds moving in a cycle back to the original sender
- Smurfing: breaking down large transactions into smaller ones to avoid detection.
- Money mules: networks of individuals used to move illicit funds.
While the pattern could be easily identified, in most cases, the transactions fit what’s considered as ‘normal’. To filter out false positives requires graph algorithms such as centrality measures, community detection and graph traversal. In addition, illicit behaviours often carry additional characteristics that could be further identified using unsupervised machine learning methods such as DBScan. Community detection can enable us to look beyond the behaviour of individuals to behaviour of groups, such as money mules.
Enhanced entity resolution
Often in conferences, AML practitioners emphasis the importance of Know Your Customer (KYC). In money laundering, fraudsters often use multiple identities, shell companies to hide their tracks. Entity resolution in AML resolves where multiple, disparate data records are referencing the same real-world entity, despite differences or inconsistencies across various sources.
With graph database, entity resolution becomes much easier. For example, identities sharing the same address, similar emails; companies sharing the same group of directors, contact details, etc are mapped to provide a dynamic holistic picture of relationships.
Discover how Napier AI can give you return on investment from day one, without any prior AML expertise in your team
Photo by Shubham Dhage on Unsplash
