Banking on tech: Lenders keep an eagle 'AI' on mule accounts
Nov 14, 2024
Synopsis
Previously banks would rely on complaints raised by customers to check fraudulent activities, but now they are trying to set up parameters which can detect mule accounts even before they get used for large value transactions
Banks are using artificial intelligence to fight the growing menace of mule accounts, deploying AI tools to track suspicious patterns like dormant accounts receiving large credits or multiple accounts showing identical recurring transactions simultaneously.
Financial technology companies such as Bureau, Clari5 and Datasutram are helping banks deploy AI-based fraud detection systems that can detect issues like mule accounts – bank accounts criminals use to launder money, often without the original account holder’s knowledge – in real time, rather than waiting for defrauded customers to raise a complaint.
Banks are using AI tools to constantly analyse their savings and current accounts. Whenever there is an anomaly spotted by these software systems, red flags are raised for subsequent investigation, industry insiders told ET.
Banks are trying to get ahead of fraudsters with this change in fraud detection strategy from being reactive to proactive, even as fraud detection has become a regulatory and government mandate amid increasing frauds.
Some of the large banks have been reporting fraudulent transactions worth Rs 400-500 crore in a month, industry insiders said.
“Previously, mule accounts would get discovered only either at the transaction stage or after the transaction when a customer filed a complaint. That would often lead to longer timelines in tracing stolen money,” said Ranjan R Reddy, founder and chief executive officer of Bureau Inc, which provides fraud prevention and identity verification solutions for businesses. “However, banks now are checking for potential mule accounts at the onboarding stage itself, and regularly monitoring their existing account base.”
While these are still early days, industry insiders said some of the solutions have helped predict mule accounts at an accuracy of around 80%.
Bengaluru-based Clari5 is working with a large public sector bank. scouring through its account base to detect possible mule accounts. From around two to three checks these banks would deploy in the past, Clari5 has broadened the scope of analysis to around 200 data points.
“Traditionally, banks relied on a limited set of parameters for fraud detection, focused on debit transactions… We are leveraging our own mule detection models to analyse over 200 attributes, including customer demographics, geo-location and device intelligence,” said Balaji Suryanarayana, chief operating officer of Clari5.
The firm’s models have been operational at banks for around six to seven months now. While banks decide which accounts to analyse and check, every account is being daily run through these machine learning models.
Additionally, startups are using software stacks that can analyse data trends from six months to a year, which helps them create a base of most vulnerable accounts which are prone to mule attacks. Previously banks would track flagged accounts for a month to 45 days at the most.
What has also helped these startups is that budgetary allocation has gone up at these financial services companies, keeping pace with the rise in fraud attacks, thereby opening major revenue channels.
“Given the jump in digital payment transactions, frauds have gone up (and) so has the budget at banks to tackle these issues,” Reddy of Bureau said. “Also, these have now become board level conversations, which means the management is very serious about implementing these solutions,” he said.
[The Economic Times]
