Posted by Checkbook on Aug 08, 2019

Checkbook's Fraud and Anomaly Detection

Do you know how we prevented millions in fraud?

As consumers and businesses are slowly transitioning into various forms of non-cash payment, fraud is becoming a lucrative – and easily-undermined – venture. According to a study by the Association for Financial Professionals, paper checks are the most targeted payment method for fraudulent claims and transactions, with at least 71% of businesses annually experiencing actual or attempted check fraud. Additionally, despite the existence of mobile-based check fraud, many still opt toward a safer alternative of Digital Checks.

However, in the world of big data, sharp practices can be dealt with hybrid processes that involve creating algorithms that efficiently organize and highlight suspicious transactions or claims. This method salvages thousands of hours of manual search time and allocates it to better customer service, maintaining higher security measures, and an overall smoother experience for our Checkbook patrons.

Every time an individual or business sends a transaction, our automated Checkbook pipeline constructs an estimated distribution of transaction history, utilizing a highly powerful statistical tool, bootstrapping sampling. Bootstrapping is a metric that relies on random sampling with replacement – in short, it allows assigning measures of accuracy to sample estimates (in our case, whether or not the check sent is legitimate). We construct a distribution scale of the individual’s average Digital Check volume, the average time when checks are being sent, and the time between checks sent. With the given data, a calculated assumption is made to determine a Digital Check’s authenticity.

Thus, with every new transaction made by the consumer and/or business, we double-check the transaction against the created distributions, and based on what it computed, whether there is a probability that the check sent is of a similar volume or greater, sent at a usual time, or at the same frequency from their last check. Anthony Sanchez, Head of Operations and Fraud Detection at Checkbook, comments, “Because our fraud detection was largely developed in-house, we were able to build a refined system of pattern-detection from the ground-up, and were able to accurately quantify variables that augmented traditional techniques and made our fraud detection second to none.”

When developing efforts towards fraud prevention and detection, we discovered that data has a tendency to become “noisy” when there are too many factors taken into account. Thus, we itemized each possible factor and calculated a risk – an anomaly score – using an Isolation Forest for quantitative factors and various clustering techniques for categorical factors. All these scores are assembled into a cumulative product, and a verdict is made. The most difficult fraud to detect are those cases where a malicious user undermines transactions into a larger account in an unobtrusive manner and slowly proceeds to take advantage of the compromised account. By using the ensembling technique, it ensures that we take into account regressive fraud.

Checkbook additionally employs device fingerprinting, allowing our team to know the approximate location where an individual or business are sending most of their Digital Checks, the hardware used (Mac, Windows Machine, etc.), and what browser being utilized. With a higher than 99.9% accuracy window, we are able to locate fraudulent claims or transactions when it is being sent from an unknown device, and block the transaction. Checkbook also deploys risk detection, where we discern blacklisted IP addresses known to be malicious and blocked these from our end.

With a combination of statistics, probability, and machine learning, Checkbook remains on the cutting-edge of fraud and anomaly detection.

Related Posts

Understand Same Day ACH

Learn about how Checkbook uses the ACH Network to provide simple and secure ACH payment solutions.

To Avoid Fraud, Avoid the Paper-Trail

A paper-trail not only makes a business susceptible to human error, but is also difficult to store and filter through in times of need.

How Small Business Can Leverage Digital Checks to Get Paid Faster

The faster, simpler, and greener payment solution for small businesses: Digital Checks.

Ready to get started?

Contact us or create an account