Digital fraud is one of the most pressing issues facing the banking sector, with fraudsters deploying a vast array of methods to target financial institutions (FIs) and their customers. Loan fraud complaints totaled 17,349 in Q1 2020 compared to 4,416 complaints during the same period in 2015, a 752 percent increase. Instances of credit card fraud jumped from 17,236 in Q1 2015 to 45,120 during the same quarter last year, a 436 percent increase. Financial fraud as a whole rose by 434 percent during this five-year period, including a 104 percent jump between Q1 2019 and Q1 2020.
These trends clearly illustrate that many banks’ existing fraud prevention solutions are not quite up to snuff. The ongoing pandemic has worsened the fraud problem, with Americans filing 184,000 COVID-19-related fraud reports through August 2020. Many banks are scrambling to replace their legacy fraud prevention systems with more advanced solutions bolstered by artificial intelligence (AI) and machine learning (ML) technology as they work to stem the rising tide of schemes.
The following Deep Dive explores traditional fraud prevention systems’ shortcomings and how they have failed to prevent a surge in fraud. It also sheds light on how AI- and ML-driven systems could turn the tide and help FIs fight back.
The Drawbacks Of Traditional Fraud Prevention Systems
FIs and businesses have traditionally relied on two techniques to counter fraud: human analytics teams monitoring for suspicious activity and static rules-based fraud prevention systems that attempt to flag fraudulent transactions. The biggest problem with the former method is the amount of time it takes to review transactions for fraud, with one study finding that a single analyst working for a business that conducts a thousand transactions per day would require 50 hours to review each one. Such a business would require seven employees to comb through its transactions within an eight-hour workday.
Many businesses attempt to streamline their fraud review processes using automated rules-based systems, which relieve analysts’ workloads but present their own challenges. False positives — in which legitimate customers are flagged as bad actors — are especially problematic when using these systems. One study found that up to 15 percent of card-not-present (CNP) transactions were incorrectly flagged as fraudulent, which ultimately prevented legitimate customers from completing their orders. Some businesses attempt to augment their rules-based systems with manual reviews to reduce the number of false positives, but this often leads to lengthy and resource-intensive processes. As much as 50 percent of all eCommerce transactions require manual review when attempting to combine these two solutions, for example.
FIs and merchants looking for more efficient and less cumbersome fraud prevention systems must look into leveraging advanced technologies instead. AI and ML solutions could well be the key to enabling more effective cybercrime detection methods.
How AI And ML Improve The Fraud Prevention Process
AI and ML are widely recognized for their risk management and customer service benefits as well as for their fraud-fighting capabilities. A recent PYMNTS study found that 64 percent of FIs that use AI said that doing so increased customer satisfaction, while another 64 percent said their charge-off rates have declined since implementing such systems. Fraud fighting is the primary goal of such initiatives, however, with 63 percent of FIs believing that AI can effectively prevent fraud and 80 percent saying that it is critical in preventing fraudulent payments.
Various case studies support these assertions. Denmark-based Danske Bank reported a 50 percent increase in detected fraud attempts after implementing an ML-based system, for example. It also achieved a 60 percent reduction in false positives, and it is expected to reach 80 percent over time as the smart system learns from its experiences and refines its techniques. This learning component sets AI and ML systems apart from their rules-based counterparts, and it is particularly useful for finding patterns in transactions that human analysts would likely miss. These could include fraud attempts coming from IP addresses in a single geographic region, for example, indicating that a group of coordinated hackers is working to breach a business’s defenses.
False-positive reduction is also a key area of improvement, as it can lead to increases in customer satisfaction. Customers are much likelier to continue patronizing banks and businesses that do not intermittently reject their transactions while also providing security methods that continue to keep their details safe.
Fraud may be here to stay, but AI and ML security systems could do wonders to blunt its impacts. This in turn could help make consumers’ experiences more secure — and more convenient — than ever.