Fraud perpetrators are continually evolving their tactics to evade detection and one that’s causing serious problems is called synthetic identity theft, where real and fictitious ID data are melded. In the US alone, this form of fraud for unsecured credit products is expected to reach US$2.4 billion in 2023, according to Aite Group research.
Money laundering is another financial crime that can be extremely difficult to detect. The International Consortium of Investigative Journalism’s FinCEN Files uncovered more than US$2 trillion in suspected dirty money had moved through the global financial system between 2000 and 2017. More than $200 million in transactions flowed through Australian banks.
In general, sophisticated layering techniques used to mask parties, with numerous small transactions moved through a maze of agents, companies and financial institutions, make it immensely challenging for investigators to follow the money trail.
Global banks process millions of transactions daily that involve tens of millions of parties – making fighting financial crime a very expensive undertaking – which includes different languages, disparate transaction formats and unstructured data. Automation helps, with AI, machine learning and natural language processing (NLP) all playing a role.
But these techniques fail to account for relationships in the underlying networks of parties and payment chains. With traditional forensic methods, data is stored as rows and columns in tables. Connections and patterns that could provide critical clues aren’t detected.
Instead, investigators are now turning to graph data platforms to detect financial crime. Graph databases store rich connections in the complex networks of people, places, institutions, behaviours, and times. Data is stored as linked nodes and relationships such as “registered at” or “transacted with”.
It’s a much more effective way of identifying chains and rings of people, generating “guilty by association” scores based on the quantity, quality and distance of a party’s relationships with suspicious entities. Algorithms create scores that are based on paths and hops from start to end points. It also becomes more powerful over time – after verifying the pattern of one fraud ring, a similarity algorithm can also be used to identify other potential fraud participants and rings.
As well as detecting and preventing financial crime, graph data platforms hold further benefits for financial institutions.
Bank security teams face an overabundance of data that it is impossible to fully analyse with available resources. It’s very manual, time-intensive work. With multiple potential alerts to investigate, security analysts end up relying on “gut feeling” to pick the ones to pursue, since they can’t fully (manually) investigate all of them.
By detecting potentially suspicious connections and activity at an early state, using graph databases enables a much more preventative and proactive approach. The technology can identify and prevent cyber-attacks before impact and make critical assets safer and more resilient. It empowers risk analysts to grant or deny requests, as well as tracing employees’ actions to detect and prevent cyber breaches and fraud.
Graph databases can be used to reengineer data flows from front-to-back to support global clients and enable rapid innovation. In private banking, clients typically hold an extreme range of asset and liability types, engaging banks in a broad range of services globally. Their connections with other beneficiaries, financial advisers and bank relationship managers are immensely complex.
This makes it very challenging to keep up with rapidly evolving global banking regulations as well as roll out product innovations. Citi has tackled this by creating a system based on graph databases that can represent myriad connections and relationships, enabling the bank to manage these relationships, handle cross-border issues and apply business rules much more efficiently. For example, it can ask questions such as “what public figures are connected to other public figures”.
Another application for graph technology is to create more robust systems for risk data aggregation and internal risk reporting. UBS needed to comply with regulations put in place to strengthen systems for risk data aggregation and internal risk reporting in the wake of the 2007 global financial crisis. The Basel Committee on Banking Supervision standard required banks to provide transparency into the data flows feeding their risk reporting.
After trying a table-based system and realising it wasn’t fit for purpose, UBS built a new tool using graph data platforms . This allowed the bank to much more easily traverse connected data and visualise relationships. Querying and reporting are now much faster and easier, and entire lineages can be exported.
The impact of even slightly better fraud detection can be significant – just a fractional percentage increase in accuracy from using graph data platforms can drive millions of dollars in savings.
By Peter Phillip, ANZ General Manager – Neo4j
This article was first published by AB+F