Visas to use advanced AI like unsupervised learning to fight fraud


Join the leaders of the Data, Analytics & Intelligent Automation Summit presented by Accenture. Register here.

The thing about cheating is that it’s constantly changing – looking at a past attack isn’t always a good indicator that the next attack will look the same or that it will target the same type of victims – and defenders need to keep changing to adjust. Visa uses artificial intelligence to analyze all transactions that go through the network and track large transaction changes as part of its fraud detection efforts, said Melissa McSherry, Visa’s senior vice president and global head of data, security and identity products at the Transform virtual conference 2021 from VentureBeat on Monday.

Visa evaluates all transactions that go through the Visa network, allowing them to define a set of behaviors that are considered “normal”. The team “constantly” updates the model’s view of history and updates the model to reflect the data fairly regularly, McSherry said.

“The fraudsters do not stand still. And they are always on the lookout for innovations, ”said McSherry in an interview with Jana Eggers, CEO of Synaptic Intelligence company Nara Logics.

The ability to see changes in the data is useful for authentication, McSherry said. A single phone and email address pair is likely associated with a legitimate transaction, especially if the same pair has been used for many transactions. The next transaction made with the couple is also likely to be tracked as legitimate. However, if that one phone number is linked to 500 email addresses, it is more likely that all of the email addresses are linked to compromised accounts, and the phone number is not linked to any real identity either.

“It is absolutely true that the data is always evolving, but we are taking the speed of the enormous amount of data we are receiving and trying to develop our perspective with it,” said McSherry.

Fraud, detect weaknesses

Everyone knows Visa has a tremendous amount of data, and McSherry was able to shed some light on how Visa uses it. Visa has been using neural networks for fraud detection since the 1990s. Eventually, the self-learning technology updated the framework of what was “normal” to detect large deviations in the model distributions. More recently, the company has been using Convolution Neural Networks (CNN) and Recurrent Neural Networks (RNN) to improve pattern recognition across the network. They not only serve as models – but also around the models to identify areas that need further examination or to highlight changes that need to be made to the model, McSherry said.

Visa uses Generative Adversarial Networks (GAN) to create virtual scammers and pit them against the anti-fraud tools to identify spots where there are gaps in their fraud detection models, McSherry said. The gaps can also be in partner-provided tools or in business logic.

“In my experience, sometimes things don’t work exactly as you think they will when you first use them,” said McSherry. New methods are used in parallel or only for monitoring until they are better understood.

Make the commitment

Integrating AI requires dedication and assertiveness, McSherry said. Visa consistently sees an increase of 20 to 30% for advanced AI methods compared to “more garden variant technologies”, but it requires high investments. Stakeholders need to stay engaged and focused as the first few attempts may not work exactly as planned. Experimentation and patience are key.

“The first and foremost thing is just to make sure that the problem itself benefits from these types of elevators,” said McSherry. “It’s just very helpful when everyone understands the value on the other side [of the implementation] is really worth a lot. “

As employees learn newer techniques, organizations can get the most out of machine learning. While it makes sense for new people with a strong AI background, Visa also gave existing employees who understood the business a chance to experiment and learn new techniques.

“I think when people who really have the business context and are proud of the quality of the product over the long term, [are combined] With a really good understanding of AI techniques, you get something very special, ”said McSherry.


VentureBeat’s mission is to be a digital marketplace for tech decision makers to gain knowledge of transformative technologies and transactions. Our website provides essential information on data technologies and strategies to help you run your organization. We invite you to become a member of our community to gain access:

  • up-to-date information on the topics of interest to you
  • our newsletters
  • protected thought leader content and discounted access to our valuable events, such as Transform 2021: Learn more
  • Network functions and more

become a member


Leave A Reply