Implementing AI and machine learning tools to detect and prevent application fraud

Implementing AI and machine learning tools to detect and prevent application fraud

By Ido Lustig, Chief Risk Officer, BlueVine

Meet Ido Lustig, at X-Tech 2019: Financial Services and Technology (Get 15% discount on the Convention using presenter code: XTECH33)

Can you please tell the Risk Insights readers a little bit about yourself, your experiences and what your current professional focus is?

Before launching my career in Risk management, I was an educator. I’ve spent roughly 10 years working with underprivileged youth and adults, helping them build a better future for themselves. It was highly rewarding work which I enjoyed to a great extent.

Eventually, I moved on to my other passion – data. I started as a Fraud Analyst at PayPal, quickly got promoted to manage the team and later the department of Data Science and Advanced Features Creation for all Risk Fraud models. Towards the end of my PayPal days I was lucky to be given the opportunity to work on some credit related features, where I got exposed to the fascinating (and much more complicated) world of Credit Risk.

I subsequently joined BlueVine, a rising star in the lending industry. I am the company’s Chief Risk Officer, in charge of all fraud prevention, credit underwriting, portfolio management, data science & analytics and the back-office teams. We work together to provide small business owners easy access to capital and to build new features and products that would address all SMB’s financial needs.

What, for you, are the benefits of attending a conference like the ‘X-Tech 2019 Convention’? What can attendees expect to learn from your session?

It’s always great to meet with industry leaders for a couple of days to share knowledge, brainstorm, and get updated on the latest trends in the industry. I’m sure X-Tech will provide many opportunities for me and the other participants to learn about new trends, methodologies and vendors and explore potential collaborations.

I look forward to talking about the various ways AI and ML can be used for detecting fraudulent applications and show a few cool examples of how we do that at BlueVine. I’m also excited to share my thoughts on how we can improve, as in industry in protecting ourselves against fraud.

In your opinion, how can AI and Machine learning tools be used to detect and prevent fraud application?

Fighting fraud is a never-ending battle. Fraudsters are smart and hardworking criminals continually evolve. There are many types of fraud, the mostly known ones are account takeover, identity theft, documents forging and the new and raising threat of synthetic identities.

To fight these, we need to always learn the latest trends, evolve and reinvent our solutions. AI can help us in all of these, mainly if we are able to build self-training models that learn and evolve autonomously. With these capabilities, we can leverage image recognition techniques to detect doctored documents, anomaly detection systems to spot velocity attacks and other abnormalities, graph-based tools to prevent repeat offenders, and so on.

AI can also help us detect gaps in our processes and methodologies as big-data based models can be built to be bias-free and point out gaps that human analysts may be missing.

On top of all these, we have just started, as an industry, to really leverage AI and ML for fraud-prevention and I expect the role of these advanced techniques to grow even more within the next few years.

What are the benefits and implications of data sharing between banks and credit bureaus?

Data is a powerful tool. When used in the right way it can help us do good. With more data being shared we can make sure people get access to capital, financial tools and services. But with great power comes great responsibility. We need to make sure data is shared in a thoughtful manner, by making sure the systems we build share only information that is needed and not beyond. This is particularly important when sharing financial data.

Could you provide insights on using machine learning tools to access data?

Machine Learning allows us to answer questions, using data, in an accurate and scalable way. These Answers are new data points we can leverage in our decision-making process.

Imagine a perfect world in which we would know all the questions we need to get answers for in order to make the right decision, be it fraud or credit related. These questions can be the time a business has been operating, their annual revenue, net margin, expected earnings and many others. With an accurate answer to each of these questions, one would be able to make the right decision, no false positives or false negatives. For a Risk Officer, this is the perfect world we strive to reach. AI and ML are the tools through which we can get these answers, by deploying as accurate models and algorithms as possible.

At BlueVine, we use ML and AI to answer hundreds of questions as part of our on boarding and portfolio management processes. Examples for such ML based models (or answers…) are automated industry identification, calculation of probability of default, fake documents identification, financial data analysis and projections and many (many) more.

How do you see the technology and innovation space evolving in the next 6-12 months?

One of the problems the industry is dealing with these days is real-time loans stacking. Fraudsters apply to multiple alternative lenders in parallel, get approved by some/most and then draw funds in parallel from all. There is no current real-time data sharing program that allows lenders to share, in real time, and while maintaining the highest standards of privacy and regulatory requirements, the fact that the client one is about to fund just got funded by another lender(s).

Another urgent problem is synthetic identities, which is many times related to the above gap – using online identities that have been built for months if not years, fraudsters are able to perform real time loans stacking for larger and larger sums, as their synthetic identities seem to be clean and solid.

I believe a main focus in the upcoming year is going to be data sharing for quicker and better fraud prevention. Using real time and anonymized data sharing I expect the alternative lending industry to better handle synthetic identities and real time loans stacking, two of the most urgent issues we are dealing with in the industry today. Some work has started, and we’re definitely moving in the right direction, but more is to come.

Meet Ido Lustig at…

Sign up for your free account:


Don’t miss out on…

X-Tech 2019