The intelligent way to detect fraud

We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 – 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!

Artificial intelligence (AI)  is best when tasked with finding hidden patterns within large datasets. So, it’s no wonder that one of the leading AI applications at the moment is detecting fraud, which is all about hiding patterns within large datasets.

As a practical matter, combatting fraud by any means necessary contributes more to the bottom line than just about any other business function except sales and marketing. At the same time, it enhances trust and brand loyalty (since no one likes getting ripped off) and strikes a blow at the lawlessness that has characterized the nature of digital transactions for so long.

Rapid uptake

According to risk management software developer, Provenir, fraud prevention is at the top of the list of reasons why Europe’s leading fintech providers have deployed AI. More than 90% of executives surveyed recently cited fraud prevention as a key driver in the adoption of AI-enabled risk detection in the past year. Perhaps the primary advantage that AI brings to fraud detection is its ability to learn as it operates. Fraud is in a constant state of evolution, so accessing real-time data and applying it to the latest defensive measures in a fully automated fashion is a major step forward compared to previous generations of financial security software.

This is probably the main reason why financial service providers like Visa are seeing record-low instances of fraud, despite the dramatic rise of on-line activity during the pandemic. 

In a recent interview with Datanami, Dustin White, Visa’s chief risk data officer, said that losses due to fraud are down to about seven cents per $100 despite more than 2 million fraud attempts per day. Over the past five years, the company has poured $500 million into AI-driven detection and prevention capabilities, particularly its Visa Advanced Authorization (VAA) scoring platform that employs machine learning (ML) and other tools to evaluate transactions in as little as 300 milliseconds. More recently, the company has launched a behavioral analytics engine to thwart account theft and bot attacks.

As with most things it touches, AI’s biggest contribution to fraud detections is operational efficiency. Görkem Gençer, at analytics firm AIMultiple, noted recently how intelligence upsets the basic equation that has driven fraud in the digital age: that is, the ease at which it can be carried out and the profit it generates versus the difficulty of identifying and preventing it from using conventional means. 

By using multiple AI models for everything from test definition and data cleansing to data extraction and analysis, organizations can increase the accuracy of their detection capabilities and do so at faster speeds and on more precise, granular levels of data activity.

Hidden figures

A close look at the many ways in which fraud can be conducted with just a single document illustrates the way in which AI allows the enterprise to up its game in prevention and detection. A new platform called Inscribe specializes in document-level fraud with its  tools that are designed to spot alterations to JPEG or PDF forms, as well as format inconsistencies that might otherwise go undetected. 

In addition, it identifies the digital trails that can be used to verify legitimate users or track down wrongdoers. Most documents also contain numerous non-identity data points, such as income, revenues and assets, that can be used to separate fact from fiction. By automating all of this analysis under an intelligent framework, Inscribe allows organizations to incorporate fraud detection more easily into the application and approval process, not just to increase the fraud detection rate but to speed up loan fulfillment and boost profitability.

 Technology is neither good nor bad, of course, so the same basic intelligence that can be used to combat fraud can also make it more effective. This is why being a laggard in AI can be highly detrimental when it comes to fraud. Before long, manual means of protection will likely be overwhelmed by the intelligent modes of attack entering the criminal mainstream. While a sophisticated defense framework will not only offer superior protection but will be able to maintain high levels of performance with relatively little oversight.

We will probably never achieve full protection against fraud, but an intelligent strategy will at least bring the cost of protection and the rate of failure down to acceptable levels.

Originally appeared on: TheSpuzz