Banks Explore Graph Technology to Tackle Data Complexity for Compliance

Banks are exploring graph technology, a method of data storage that shows the different relationships between data points, for a number of use cases, including fraud detection, anti-money laundering, and regulatory compliance.

Knowledge graphs offer an alternative to siloed, disconnected data stored in relational databases, as they represent domains of knowledge, connecting information in a systematic way. Knowledge graphs encode information in a network of nodes, like a data map set to specific standards, rather than tables of rows or columns.

One way to understand it is to compare a spreadsheet to the internet. In a spreadsheet, there is a column and a row that meet and give a datapoint. Therefore, that datapoint is standardized and defined by two things—the column and the row. On the other hand, every piece of data on the web can be mapped to an unlimited number of standards. For example, one search can pull up hundreds of thousands of relevant results because each URL that appears from the search contains data that shares the same set of standards as the search.

Some banks have begun exploring this technology to catch potentially suspicious trading behavior. Detecting insider trading is not a straightforward task: While many banks have security systems in place to catch individual trades that fall afoul of regulations, often a trade when seen alone may not appear to break any rules. Rather, there is a wider pattern around trading and communications data that needs to be investigated. 

“Seeing those links, understanding those connections, and making them visible just by moving from a table into a network structure or graph structure, it is fascinatingly powerful,” says Rik Van Bruggen, EMEA vice president at Neo4J, a provider of graph database technology. 

A number of Neo4J’s clients have started exposing “their own internal information about clients in a knowledge graph interface, but they also include external information sources, like news feeds, geographical information or personal anecdotes, emails, whatever it may be. All of these things get pulled together to create this knowledge graph,” he says.

Mapping Through Graphs

Wells Fargo’s chief innovation officer, Lisa Frazier, spoke at the Sibos conference last month about how the bank is using knowledge graph technology across multiple datasets “to simplify and understand data relationships across all our data silos.”

Once the data is mapped into a knowledge graph, it starts to give a “fuller picture” of customer relationships to the bank, Frazier said. This means taking every disparate dataset and essentially filing it into one system.

Frazier said the bank’s first use case for knowledge graphs was for data privacy regulations. A customer of Wells Fargo worked with the bank on the new laws rolling out in California. The state will begin enforcing the California Consumer Privacy Act, which went into effect on January 1, 2020. Similar to the General Data Protection Regulation in Europe, the law recognizes the right of consumers to know what of their personal information is being collected by a company, and gives them the ability to opt out of data collection. Under the law, the state’s attorney general can fine companies up to $7,500 for each violation.

ING is another bank that is currently using graph technology. It is trying to identify suspicious activity within transactional data. The bank’s data analytics platform, Hunter, is designed to analyze 14 billion transactions in a matter of seconds. The solution has been fully operational from early this year, but the bank is continuing to develop its analytics models and tools for its financial crime investigations teams.

Annerie Vreugdenhil, chief innovation officer of wholesale banking at ING, recently told WatersTechnology that new advancements in analytics and graph databases are helping banks such as ING keep pace with increasingly sophisticated financial criminals. She said criminals often move their operations to different territories where they believe they will have less regulatory scrutiny, or scrutiny from institutions. In response, banks need to be able to follow and monitor these bad actors’ movements on a global scale.

Communications records, such as emails, are an important source of data for building graphs around fraud. David Haines, CTO at regulatory compliance vendor SteelEye, says knowledge graphs on its platform can give users a visual cue around who a person could be communicating with. “It is almost like six degrees of Kevin Bacon,” he says.

The idea of six degrees of separation is that one person is only six social connections away from everyone in the world. From a compliance perspective in the financial markets, with graph technology, an investigator would have the ability to look at the inner circle of a suspicious actor and identify who they were in regular contact with.

Not Exactly New

Knowledge graph technology is not new, though in recent years, Google popularized its use with its search engine. The knowledge graph pulls together and links people, places, and facts to create interconnected search results. The aim of the graph technology in that context is to make each search more accurate and relevant to the user.

“My dad was in the computer science industry in the ’70s and ’80s,” says Van Bruggen from Neo4j. “When I first told my dad about going to work for Neo4J and I tried to explain to him [about graph technology], he was literally saying, ‘This is exactly what I was doing.’ And he is right. On the mainframe we had things like navigational databases, network databases—it was exactly the same thing.”

What is new now, he says, is the greater accessibility of the technology. “I can run my mainframe-style database on my phone if I wanted to, or on a Raspberry Pi. I am making it accessible to such a wide variety of different people, organizations, and use cases,” he says.

Michael Atkin, co-founder of the Enterprise Knowledge Graph Foundation (EKG), which was established to promote the use of the technology, says he knows of three tier-1 banks that are using knowledge graph technology. However, due to structural challenges and lack of skills, all of them are still only in the test phase of the technology. 

EKG has a model to measure knowledge graph competency in companies, graded on a range of 0 to 5—from no knowledge of the technology at all to full integration with a supply chain connected and at the stage where a company can add artificial intelligence and automation to the graph. “In the finance industry, we average at about 1.7 right now,” Atkin says. Financial services is lagging behind more science-led industries in this area. For comparison, the US space agency, the National Aeronautics and Space Administration, or NASA, is at level four.

While few banks have live-production knowledge graphs currently, Atkin says there is wide interest in the technology. “Every big bank that I know of is experimenting with this,” he says.