Enabling machine learning to help combat fraud

The combination of trained fraud experts and machine learning promise a best way forward for financial institutions in the battle against fraud, in today's omni-channel experience

Machine Learning Fraud

While unsupervised models might be the goal in the long term, supervised and semi-supervised models in which trained fraud experts provide input to better train and fine tune their ML models are critical to creating a fraud prevention system that can proactively identify fraudulent activity while reducing the incidence of false positives that can negatively impact the customer experience

A recent study from iovation has centred on the movement towards using machine learning (ML) analytics for improving fraud mitigation and customer experience.

The findings showed that while 68% of financial institutions (FIs) cite ML analytics as a high priority investment over the next few years the path to adoption will not be without its challenges. iovation and Aite Group will be hosting a webinar at 10AM PST on November 9th to discuss the results of this survey.

It’s clear that the threat environment continues to escalate, and effective fraud prevention is an increasingly competitive issue for FIs. Organised crime rings, armed with billions of stolen data records, are targeting the financial services industry with sophisticated card fraud, application fraud, account take-over (ATO) attacks, wholesale ATO and the spectre of faster payments, all of which were cited among the top pain points for FIs today in the Aite research.

>See also: Man versus machine learning: How to beat the fraudsters

“What this study highlighted is that those who are early adopters of advanced machine learning analytics will be able to greatly reduce fraud while also improving the customer experience, giving those FIs a decided edge over their competitors who lag in these advancements,” said Julie Conroy, research director for Aite Group’s Retail Banking & Payments practice and author of the report. “Data is the new currency, and creating intelligence from data at scale requires machine learning technology.”

With the omni-channel approach to consumer engagement there are several points of interaction (ATMs, call centers, email, etc.), for hackers to gain access to personal data. In fact, cross-channel fraud was mentioned throughout the interviews with fraud and analytics executives as a key attack vector. All of this points to why FIs must embrace machine learning analytics to spot patterns and holistically prevent fraud at each point of contact with stronger yet more convenient authentication measures, which will ultimately improve the customer experience.

“Hackers see FIs as the perfect target for fraud. And, if you pile on the intense pressure to make the banking experience easier and more frictionless for customers it’s clear that FIs need a solution that will make significant improvements to their fraud fighting approach,” said Eddie Glenn, product marketing manager for iovation.

“We are seeing FIs turn to machine learning analytics as the foundation for a broader solution. Leveraging the vast amount of customer data at their disposal and applying advanced, machine learning techniques, FIs can create insights that allow them to better predict new fraud types they have not seen before as well as implement more contextual authentication measures that can understand the relative risk of a given transaction to ensure a more seamless experience for retail bank consumers.”

>See also: The rise and rise of intelligent machines

One of the misconceptions about machine learning is the perception that ‘unsupervised models” in which sophisticated algorithms alone will become the prevalent model to identify and reduce FI’s exposure to fraud will be the ultimate arbiter of whether a transaction is potentially fraudulent.

However, according to the report, the majority of the installations of ML among the FIs interviewed use some type of combination of scoring generated by ML and highly trained fraud experts to continually refine and improve their models. In fact, 50% of respondents are using entirely supervised techniques or primarily supervised.

While unsupervised models might be the goal in the long term, supervised and semi-supervised models in which trained fraud experts provide input to better train and fine tune their ML models are critical to creating a fraud prevention system that can proactively identify fraudulent activity while reducing the incidence of false positives that can negatively impact the customer experience.

Comments (0)