For tech-savvy fraudsters, the arrival of the internet made it relatively easy to hide money. However, Prag Sharma and Lorcan Coyle of UCD’s school of computer science and informatics are planning to join the teams striving to make it a lot more difficult.
The partners have taken the techniques used in social network analysis and applied them to identifying fraudulent behaviour in real time. Typical customers for their EgoNav product, due for launch early next year, will include banks and other financial institutions.
“Fraud is a massive, complex problem and a huge business cost,” says Sharma. “Many companies have large customer datasets with network structure – for example financial transaction datasets or mobile phone networks. Due to their sheer size and complexity, it is often difficult to understand the relationships within the datasets.
“As a result it can take weeks for companies to become aware of fraudulent transactions. This is a real blind spot that can be exploited.”
EgoNav has been developed at the Clique Research Cluster (Science Foundation Ireland’s graph and network analysis cluster initiative) and its commercialisation supported by funding of €200,000 from Enterprise Ireland.
The partners are planning to sell the product directly and to partner with organisations providing banking and financial consultancy services.
EgoNav works by identifying recurring patterns and anomalous activity in large datasets and allows companies to respond to suspicious customer behaviour very quickly. “EgoNav performs a parallel analysis of all customers in real-time and the algorithms constantly evolve to follow the shifting customer behaviours that typically underlie fraud,” Sharma says.
“The platform includes a visualisation component which can be used to interactively browse a network dataset to highlight typical classes of customers and to identify and examine unusual activity. This information is then made available to reporting engines or alarm functions, which ensures a fast response to exceptional behaviour.
By applying network analytics to the interpretation of complex behaviour, EgoNav offers a more human-intuitive view of data that is more usually viewed in tabular form. This enables ordinary business users to perform the analysis rather than dedicated data scientists,” Sharma says.
While the current focus is on using EgoNav to detect fraud in a financial services environment, Sharma adds that the core technology can also be applied in other ways.
“It would also be useful in the area of precise customer segmentation.
“Today most companies have a lot of data about their customers, but they struggle to use it effectively. Our system would allow them to use what they know to be much more specific about whom they target. For example, it may not be the first layer of contact they need but the second, friends of friends perhaps, who are more likely to be in the group a company needs to reach.”