Keeping Up With Fraudsters: A Month Isn’t Enough

As the Government of Canada campaigns for improved fraud prevention and awareness this month, I’d like to do my part as a fellow Canadian, and shed some light on why payments need to stay a step (or more) ahead of fraudsters, today more than ever.

Growing variety of fraud types in a dynamically-changing industry

Fraudsters have evolved with the times. They no longer use only one tactic to dupe their victims and steal money. They have built up an arsenal of different approaches to scam people based on social engineering, and they are after more than just your money. They listen for cues when you talk, or in how you act, to provide the right incentive for you to share your password, provide access to your computer, or push a credit transfer to their accounts.

They commit their crimes through the internet on dating sites and social media, through malware that steals your credentials, by hiring you for a new somewhat-believable dream job, or convincing you they are the government and you must send them money for overdue taxes. They use your stolen identity to secure credit cards and loans in your name, worth more than the value of your savings account. With all the variety when it comes to fraud, it’s hard for the consumer to keep up.

While the campaign for greater fraud awareness shines a light to the issue, the reality is that fraudsters are constantly evolving and changing their patterns. So, that begs the question: is one month really enough for banks, businesses and consumers to fathom the various scams out there and fight back? I think the answer is no. This is where machine learning comes in.

A primer on Machine Learning

Some background on Machine Learning. In payments fraud prevention, Machine Learning (ML) uses massive amounts of transaction data – both good and bad – that has been accumulated over several years to ‘learn’ the difference between legitimate and fraudulent transactions and help payment institutions identify and stop transactions before a consumer’s account is emptied.

Within ML, there is supervised learning, where humans help the computer understand what is good and bad, and unsupervised learning, where the computer is given basic instructions and works within these rules.

Most banks, processors, and retailers are currently focusing on supervised ML, as it is easier and faster to build with existing data, stops fraud, and makes it easier to explain how it was built – a requirement by regulators/governments who want this information for the sake of transparency. It also leverages the existing knowledge of in-house fraud prevention experts to define the initial rules.

However, many are now turning to unsupervised machine learning as it promises to build upon that human knowledge and learn to identify new types of fraudulent transactions and behavior, much faster than its human overlords. Although still relatively nascent, it has made great strides in recent years. Expect to hear more about it in the near future.

Consumer education is critical

Back to Fraud Prevention month. As banks, payment providers and retailers have improved their fraud controls, fraudsters are finding it harder to steal from them and scammers have realized that the weakest link is now the end-citizen. This is driving a massive push towards social engineering scams, where the customer is duped into legitimately sending money to the bad guys – these are known as authorized push payments. To that end, fraudsters have created hundreds if not thousands of different types of scams with the goal of finding the scam that makes the consumer click and eventually send them money! By constantly changing their approach it is hard for banks to stay a step ahead.

Machine learning, on the other hand, might just be able to identify those hundreds of variations of scams and stop fraudsters before they cause problems for consumers. Even with the ongoing deluge of data and new threats across institutions, machine learning can learn and identify new risks long before they are mainstream and in the media.

As our global economy moves toward digital forms of payments – both immediate and real-time, with 726 billion transactions expected to be made by 2020 –  – and monumental volumes of payment intelligence data out there, machine learning is going to play a big role in solving fraud and financial crime!

Join Marc for a live webinar on March 21 at 11:30 a.m. ET, where he will address how banks can drive real value to their fraud prevention strategies with machine learning analytics if they cut through the hype in Key Trends in Payments Intelligence - Machine Learning for Fraud Prevention.