Leveraging Big Data to Revolutionize Fraud Detection

Banking is a massive, complex industry with many facets -- retail banks, credit card lenders, managed investing, risk management -- all of whom approach fraud detection and prevention differently. Credit card fraud makes up around 40% of the total problem. According to Consumer Sentinel Network, U.S. Department of Justice, the total amount of credit card fraud worldwide is $5.55 billion and on the rise.

The sheer volume of loss attributed to fraud is pressuring financial services companies to devise solutions to prevent and identify fraud, while continuing to provide a positive and customized experience for an increasingly sophisticated customer.

In order to achieve a more accurate and less intrusive fraud detection system, banks and financial service institutions are increasingly investing to perfect the algorithms and data analytics technology used to spot and combat fraud. This technology uses large volumes of data being generated at a high velocity to increase confidence and accuracy in fraud detection.

Increasing Accuracy in Fraud Detection

As a highly customer-centric industry, banks and financial service providers need to make sure they are attacking fraud strategically and not disrupting their customers’ banking experience. The key to accurate and non-disruptive fraud detection is to implement emerging technology that allows banks to gain a holistic view of customers.

This view of fraud detection uses data available from a variety of sources -- mobile data, along with social data from Facebook and Twitter -- and uses it to distinguish fraudulent activity from normal activity. For example, if a credit card customer fails to alert her card lender of her travel plans, a strategically implemented fraud detection system can enable the lender to automatically gain insight from mobile and social data that the customer is traveling and thereby reduce the incidence of false positives.

Using social data to ‘cluster’ information is another innovative approach. For example, if the cardholder’s social media contacts are tagged as coworkers and their social profiles indicate their current geographical disposition, a bank can use this information again to correlate recent charges incurred by the cardholder.

Machine Learning

An emerging and powerful aspect of big data and fraud detection is machine learning. In a nutshell, machine learning takes place when agile systems are configured to learn from one another. It discovers the patterns buried in data and learns from itself to deliver higher quality insights, helping detect fraud in real time and adapting its systems to more quickly identify fraud in the future.

Machine learning models are increasingly being used to screen financial transactions for fraud. Text mining and machine learning technologies are effectively used to combine data from suspicious transactions with related extracts from other internal and external sources (such as social networks). One immediate benefit is to effectively reduce the amount of false positive hits thrown out by the existing surveillance systems, thereby reducing the costs of manual inspection and making fraud detection more relevant and accurate.

Best Practices for Implementing Big Data

Banking and financial services institutions are increasingly aware of the benefits of leveraging big data in fraud detection, but often struggle with where to get started. The following best practices should be considered when developing a strategy:

-- Start with small and specific uses for big data: The first thing organizations should do is identify one or two business problems that can be resolved by improving fraud detection, and then dedicate the R&D resources to develop solutions. This type of ‘outcome-based thinking’ will ensure the business success of the initiative.

-- Ensure you’re working with high quality data: Take the time on the back end to ensure you are collecting the proper data and are separating the signal from the noise to allow for proper data analysis.

-- Know your regulatory environment: Understand the boundaries for using customer data and the relevant privacy laws. This will continue to be a challenge for organizations, but can be managed with the right approach.

-- Ensure IT and business units are collaborating: Implementing big data systems can be disruptive to an organization. Adopt destination-driven thinking where you and your team articulate and agree upon clear goals, and then evangelize your big data strategy across the organization to gain broad buy-in. These clear objectives will motivate teams to work through inevitable friction.

Looking Ahead

Technology will continue to advance and offer new strategies for optimizing fraud detection. Financial services organizations should ensure their internal teams are informed about trends and developments, or are partnering with experts who effectively help them stay ahead of the technology curve.

Big data allows financial services firms to greatly enhance the speed of fraud detection and prediction using massive amounts of data from a hybrid of sources: point of sale, social media, customer databases, and external sources from data vendors. The analytic results from this data are growing into a collective “fraud database” for the financial services industry, which is driving new analytical models.

In the future however, big data will help deal with the globalization of fraud itself. On a global scale criminals are developing fraud tactics and scenarios driven by data and analytics. They use data to probe for weaknesses and monitor the “success” of fraud programs they initiate. As computing resources become cheaper and faster in an internet world where a “location” is very much a virtual presence, criminal enterprises can move operations in a borderless digital world.

The challenges that the financial services industry faces with fraud have an enormous impact on customer service and the fight to lower fraud loss. Real-time analytics and machine learning built on top of a big data repository represent the solution platform for fraud detection and predictive/preventable fraud while maintaining a highly level of customer satisfaction.

Karthik Krishnamurthy is VP, Enterprise Information Management at Cognizant.