Deep Dive: Using AI To Solve The $4.2T Fraud Problem

Digital payments volume is increasing as emerging technologies enable more seamless online transaction experiences for consumers and businesses. Cybercriminals are also seeing opportunities to use these offerings to their advantage, however, now employing them to exploit vulnerabilities, siphon funds and access valuable data. 

Merchants and FIs have been turning to such solutions to help customers quickly and securely transact. However, bad actors are doing so to commit financial crimes and get away undetected. The latter successfully made off with $4.2 trillion from the worldwide economy last year, a problem that will become only larger as transfer speeds increase. 

Many organizations are thus tapping into advanced, unsupervised learning technologies — which provide opportunities to reduce financial fraud as detection becomes smarter and machine learning more robust — like artificial intelligence (AI) for assistance. The outlook for banks, businesses and consumers to remain protected could significantly improve with the systems’ prevalence. 

The following Deep Dive examines how AI technology is being deployed on the anti-fraud front lines. 

The High Costs Of Fraud 

Fraud takes many shapes and forms and comes with a hefty price tag. The increasing risk of fraud is prompting many firms to implement AI-based solutions. These solutions can analyze consumers’ personally identifiable information (PII) and transactional data, helping combat and identify irregular credit card activity for specific patterns — whether false positives or actual fraud. False positives occur regularly with traditional rule-based anti-fraud measures, as the systems tend to flag anything outside their given sets of parameters. 

False positives are considered one of the most significant barriers to businesses’ customer acquisition and retention efforts. A recent study found that 60.8 percent of digital platforms consider false positives to be key conversion process friction points, a major problem for online companies. Their ranks include digital platforms such as Amazon Business and Airbnb, meaning they must deliver user experiences that are seamless, secure and allow trustworthy customers to quickly conduct transactions. 

A firm attempting to make purchases from a new overseas supplier can trigger a fraud warning, for example, but AI solutions might compare it to a cluster of similar small to midsize businesses (SMB) accounts and further examine it before raising a flag. Digital platforms can thus create digital, data-based user profiles with AI-based systems in place, then determine if recent activity warrants cause for alarm. These solutions are likely to increase customer satisfaction and reduce the risk that banks will incorrectly flag genuine users’ accounts. 

Getting AI Security Right 

False positives can undermine fraud detection systems’ effectiveness, making it crucial that developers understand the importance of seamless, accurate data input processes. AI systems are only as effective as the input systems that support them, and should consider factors like geographic data and how it compares to traditional patterns, for example. Firms would do well to carefully analyze confirmed fraud alerts to understand how accurately their systems flag cybercrime-related events. 

Banks and businesses would also be wise to take holistic approaches to fraud. Data collected by firms’ marketing divisions might include insights on the most popular channels — online, mobile or in-person — through which consumers choose to interact. This can lead to further opportunities to enhance security at these access points. 

Fighting fraud requires several layers of protection to be effective in detecting criminals while approving trustworthy partners. Merchants that deal with live customers can use chip-enabled point-of-sale (POS) terminals, an EMV-enabled technology that can keep payment card data secured at the POS and assure customers that their data will be protected. 

Solutions like enhanced authorization for card-not-present transactions can share additional information like email addresses, IP addresses, shipping experiences and more to build more complete user profiles. Others, like EMV-3D Secure and tokenization services, are winning appeal by enabling merchants to shift liability and alleviate the need to store payment information. When combined with other measures, these tools have been proven to reduce fraud by approximately 60 percent. 

AI will be essential in the fight against fraud, especially as bad actors become savvier in their digital crime efforts. The technology must be taught how to correctly collect and interpret data to be effective, though. Such solutions are only as effective as their users and developers teach them to be.