Keshav Pingali

Katana Graph Solving Security Issues With Graph Technology

As the amount of data we generate grows, it requires more complex technology to be able to extract all the possible value from it. That is true in fraud detection and a young company is using graph technology to provide valuable insights and protection to financial services firms.

More and more of that data is unstructured and that makes it harder to interpret, but Katana Graph is using graph technology to identify relationships between entities which is modelled using a graph, cofounder and CEO Keshav Pingali said.

We use graphs all of the time, often without realize it, Mr. Pingali explained. Think of planning a flight somewhere. Airports are the nodes and the flights are the edges. Which airports can get you to your ultimate destination? That is a simple example of a graph.

Keshav Pingali

That is the principle behind fraud protection too. You identify the different participants in a network along with the relationships between them. Which ones interact with each other? Are there ones with established relationships to suspect players?

This has come about with the advent of graph AI, the ability to apply deep learning techniques on these graphs, Katana Graph pre-sales engineer Greg Steck explained. When working with financial institutions they aggregate all data channels and integrate them into a consolidated view that can be naturally modelled on a graph.

“You run algorithms on that to get a comprehensive view of the customer,” Mr. Steck explained.

The more areas you can obtain data from, the more comprehensive of an analysis you can conduct, he added. Obtaining data from some sectors can bring regulatory and security challenges, however, but in general, the more data, the better the assessment.

Since the onset of the pandemic Katana Graph has seen an uptick in fraud, as digital payments have accelerated. That means it’s all the more important to quickly identify relationships before payment-related fraud can occur. Graphs can identify fraud patterns because of their ability to illustrate complex patterns in high dimensional data; with all this new data we have now there can be many relationships added.

Consider money laundering, where money can move from one account to 10 others and from there to 10 others, before working its way back to one or two new ones. That process is tough to track using traditional rules-based systems, but with graphs they can better able to zero in on the money laundering schemes that are most likely to be happening. Remember the above airport analogy but see the accounts as nodes and the transactions as edges. This time look for anomalous links from unfamiliar communities that could suggest a high likelihood of fraud that could be in the form of shared IP addresses and multiple uses of the same email accounts.

Many industries can benefit from Katana Graph’s technology, but the company is limiting itself to a few sectors so they maximize competencies. One is medicine, where the technology can create knowledge graphs on drugs, organizations, diseases and research and combine them in one data base for hypothesis generation on new drugs.

“If you can use these ai techniques on these graphs and basically mine the medical knowledge you might be able to rule out a bunch of candidates without having to do the lab tests.” Mr. Pingali said.

The other is intrusion protection for computer networks. Katana Graph can build interaction graphs in real time. What are the normal lines of communication between certain parties? Which people are usually in their networks? Is there anybody knew? What are their usual networks?  If the system flags anomalies they can be forwarded to a human operator to investigate.

This system has another key advantage over rules-based ones.

“It is hard for rules-based systems (to do),” Mr. Steck said. “With graph AI, given examples, the system learns on its own how to do these inferences.”