Director of Product Marketing, Milena Babayev, chats online payment fraud, understanding the Science Inside the Ekata Identity Engine, fraud prevention, risk assessments, and digital identities with the Director of Field Data Science at Ekata, Sarah Strano.
Every time I make a purchase online, I often wonder how the retailer knows if I’m a real customer or a fraudster. Turns out that I’m not the only one as, according to Juniper Research, online merchants are predicted to lose $48B to online payment fraud globally this year.
Fraudsters constantly adapt their tactics to find new ways to commit online payment fraud. It’s relatively easy for someone to hack into my account and steal my name and my credit card information. But it’s much harder to mimic my behavior online and “act the part.” That’s where sophisticated fraud detection and fraud prevention comes in.
To understand how businesses predict fraudulent behaviors, I sat down with Sarah Strano, who heads up the Field Data Science team at Ekata, over virtual coffee to talk about the science of the Identity Network and how its models help customers find fraud, reduce customer friction and offer a better customer experience.
Milena Babayev: First and foremost, for the newbies here, can you tell me a bit about Ekata?
Sarah Strano: Ekata, a Mastercard company, is a global leader in digital identity verification solutions. We enable businesses across the globe to link any digital interaction to the human behind it. Our goal is to empower businesses to combat fraud, build trust, and enable frictionless transactions worldwide.
MB: How do you do that?
SS: All of our APIs and SaaS product suite are powered by the Ekata Identity Engine™, where sophisticated data science and machine learning combines two technologies, the Ekata Identity Graph and the Ekata Identity Network, to help reduce customer friction and improve fraud risk assessment and analysis.
MB: Can you tell me more about the Identity Network?
SS: The Ekata Identity Network helps you understand how a digital identity from the Identity Graph is being used in digital transactions using machine learning models, predictions, and behavioral patterns. For example, when did we see this email first in our network, how often is this email used in transactions, how many phone numbers were associated with this email in recent transactions, etc?
MB: What sets Ekata Identity Network risk models apart?
SS: Combined with our vast global training data of over 5 billion interactions, our advanced machine learning capabilities, especially when it comes to feature engineering and experimentation using our network testbed, allows us to iterate at a rapid rate. Our training data is something I am particularly proud of because it allows customers to leverage our data science resources to detect unique global signals that help identify good vs. bad users. These unique global signals are constantly fine-tuned to improve risk models that are integrated into a customers’ decisioning platform and rules engines.
MB: Why do we update our Identity Network risk models?
SS: We update risk models to deliver significant performance improvement. Risk models have proven to help identify good vs. bad transactions. Ekata may include updates to one, multiple, or no models. Models are monitored in production, and measured on an ongoing basis to help evaluate when an update is required.
MB: Do customers see any changes in distribution when new Identity Network risk models are released?
SS: Changes in distributions are normal and to be expected. We anticipate this since we do not build custom models. We update our models regularly to leverage current training data and provide the most up-to-date and performant data. We recommend observing the changes in distributions via the distribution chart functionality with the model management utility in the Pro Insight admin tool.
MB: Where are we going to see the most improvements in Ekata’s fraud risk assessment?
SS: The most exciting things you will see are real-time signals allowing us to detect behavioral anomalies almost instantaneously, time series features (of personal interest to me as someone who studied time series) and improved handling of missing inputs.
MB: What are some of the big updates in this most recent risk model?
SS: One of the new features is the “real-time” capability to detect intraday fraud attacks. No matter where in the world the behavioral element(s) are being used, Ekata Identity Network will provide you the risk indicators on them within seconds.
With just minutes left of our virtual coffee meeting, I couldn’t resist and snuck in a few non-fraud related questions:
MB: How do you spend your free time?
SS: I spend most of my free time with my kids! I also try to make time for exercise and travel as much as possible, which is even more adventurous with kids.
MB: What are you reading today?
SS: Code Breakers. It’s an amazing book about Dr. Jennifer Doudna who pioneered a lot of RNA research and technology. I am learning so much about really relevant science and the importance of teamwork when pursuing cross functional innovation.
MB: Favorite place you’ve ever visited?
SS: It’s a tough competition between Mongolia and Morocco. Both are amazing and magical places that allow for adventure and relaxation.
MB: If you weren’t heading up the Field Data Science team, what would you be doing?
SS: That’s a tough question. I think it’s between leading education and outreach initiatives through a science museum/National Science Foundation or working at a national lab on interesting technical problems involving big data.
And that’s a wrap.
At the end of the day, it’s important for organizations of all shapes and sizes to get better at risk assessment and identity verification to protect their revenue streams. This is where Ekata comes in! You need a solution you can trust to work behind the scenes, detecting fraudulent activities before they cause mayhem. You can learn more about Ekata Identity Network here.