Payments in Focus Insight that keeps you ahead of change

5 min read

How Machine Learning Enhances Fraud Prevention for Businesses

Wednesday, April 28, 2021

Fraudsters never take a break, continuing to find new ways to cause harm to businesses. Yet while they become more sophisticated in their approaches, so are the fraud prevention measures to safeguard businesses. Today, machine learning can be used to enhance fraud prevention, and it's helping businesses of all sizes avoid major losses.

Machine learning for better fraud detection

Machine learning is a form of artificial intelligence (AI) where computer software learns from the data it analyzes and its accuracy improves over time as it gathers more data inputs.

Machine learning has now been integrated into platforms to detect and predict fraud patterns for more accurate prevention. It's quickly becoming a much-needed tool to help businesses combat transaction theft.

Businesses are using machine learning to detect and predict fraud patterns for more accurate prevention. It's quickly becoming a much-needed tool to combat transaction theft.

Fraud platforms powered with machine learning look for subtle nuances in the payment data that are not readily apparent. They take large amounts of payment data and unearth hidden correlations between cardholder behavior and the likelihood that a transaction is fraudulent. As time goes on, these fraud platforms are able to make adjustments to the algorithms based on changes in the patterns they are "seeing."

For example, with COVID-19, fraud platforms with machine-learning capability had to adapt to a whole new set of transaction patterns. Suddenly in March 2020, a large amount of consumers were restricted to their homes, increasing the purchases they made online, in categories such as groceries, restaurants, and retailing.

In these scenarios, machine learning is very good at adjusting to the new patterns and can do so much faster than a human could. Traditional fraud platforms for merchants use a static set of rules to determine which transactions to accept, reject, or set aside for manual review. This approach is called rules-based detection. It's an effective approach, but it requires more time and human intervention to sort through data, adjust the rules, and perform the manual review of the transactions themselves. Machine learning cuts back on these manual processes and is better equipped to find subtle fraudulent events that may not be detected by the rules-based approach.

A comparsion: Machine-learning and rules-based fraud management

Machine-Learning
Rules-based
Detection
Finds suspicious transactions by searching correlations in the data to root out bad actors
Captures most obvious scenarios based on a set of rules; can't look for correlations
Verification
Fewer instances of asking customers to give another factor of authentication
Higher number of false positives and requests for additional information
Adjustment of rules
Algorithms are changed in real-time
Rules are updated manually; can be time consuming
Fraud Rate
Lower than a rules-based approach
Higher, due to above characteristics
Machine-Learning
Detection
Finds suspicious transactions by searching correlations in the data to root out bad actors
Verification
Fewer instances of asking customers to give another factor of authentication
Adjustment of rules
Algorithms are changed in real-time
Fraud Rate
Lower than a rules-based approach
Rules-based
Detection
Captures most obvious scenarios based on a set of rules; can't look for correlations
Verification
Higher number of false positives and requests for additional information
Adjustment of rules
Rules are updated manually; can be time consuming
Fraud Rate
Higher, due to above characteristics

On the whole, machine learning helps predict fraud with better precision than other methods. Our solution combines data from more than 68 billion Visa transactions worldwide and over 260 fraud detectors with machine learning of static- and self-learning models. For more tailored targeting, merchants can customize rules to meet unique business needs. It all adds up to maximum prediction accuracy to help prevent more fraud.

Who's at risk?

Every merchant–from the biggest multinationals to the smallest micro-merchants–is at risk of transaction fraud. Large enterprise businesses without sufficient safeguards are vulnerable. Cybercriminals look for high-value, in-demand goods that can be quickly resold. Smaller merchants are also at risk, as fraudsters typically migrate to the weakest link, those that haven't yet employed extra layers of sophisticated fraud prevention.

Today, many merchants aren't doing enough, and have left themselves highly exposed to potential fraud. And unfortunately, merchants often misunderstand the concept of card acceptance liability, which will fall to the seller unless it follows strict card-acceptance rules promulgated by the networks. Common refrains we hear from victims are, “I've been doing this for years and I've never gotten hit," and “It's never happened to us." Understandably, in volatile economic times such as during a pandemic, merchants might be all-too-eager to process large transactions, which at first glance could make a difference between a mediocre and good month sales-wise.

Yet if a sale is too good to be true, it often is. And it's the smaller merchants that are typically the least able to sustain big losses. Once fraudsters find a vulnerable merchant and a transaction goes through, they'll keep hitting the merchant until the business realizes there's something wrong. We often see single fraudulent transactions ranging from $5,000 to $50,000, which can be crippling for smaller operations.

Make machine learning part of your fraud prevention

At a minimum, every merchant needs a fraud strategy. Merchants that have sustained fraud losses often say that what turned out to be a fraudulent transaction was too good to be true. They accepted it and didn't trust their gut. But with stronger anti-fraud measures such as machine learning, they can let the vast computer power of a processing network do the hard work, allowing it to flag suspicious transactions with greater accuracy and lower false positives.

For merchants that can't integrate machine learning at this time, we urge them to use other advanced fraud technology such as 3D Secure 2, which uses biometrics and other methods for quick, smooth authentication on any device. 3D Secure raises the security of an online transaction to the level of a face-to-face transaction at the point of sale. Also, merchants should notify people picking up the goods in person that the transaction will be processed as a face-to-face transaction, which usually stops the fraudsters in their tracks.

Keeping up with the leading fraud prevention technology is protection for businesses of all sizes. After all, fraudsters are constantly refining their tactics. Shouldn't you?