Real-time Fraud Detection using Deep Learning

Duration: 25 mins
Agniraj Baikani
Software Engineer, PayPal

Traditional fraud protection methods in the Fintech industry have been rule-based, where a human defines the rules. However, this is not scalable due to latency and with new fraud patterns emerging every day. Also, merely classifying transactions as anomaly is not enough as we want to gain a better insight into the root cause(s) of the anomaly.

To solve this, we can use deep learning techniques to reconstruct normal transactions to find anomalies from the normal patterns in the transactions. On detection of an anomaly, our model traces the most influential variable(s) that caused this anomaly.

In this session, we will give a brief overview of deep learning, anomaly detection and how to trace the root cause(s) of the anomaly.

You may also be interested in

25 mins
Designers + Developers = Best Friends Forever?

How is the relationship between your design team and your development team? Is it highly functional? Or 'just professional'? Maybe...

50 mins
All That Glitters Ain't Gold

Let’s use Kafka, everywhere! Let’s try event driven architecture! How about Rust for this service? Let’s use Elixir for this!...

25 mins
Dealing with Noobs: How to Work with Non-Technical People

The secret to being the hero of your story? Helping others be the hero of theirs. Take a business problem you...

25 mins
How Non-violent Communication Can Help Keep the Peace on your Team

Non-violent communication will help you communicate with your coworkers in a manner that enables productivity and helps you understand how...

180 mins
Modern Software Development

Our industry never stops changing, but sometimes those changes are trivial and fluffy and we can ignore them. Sometimes they...

50 mins
Definition of Ready & Done - A Guide to Achieving Predictability

Delivering software often takes longer than we anticipate. Why is that? Part of the reason is not understanding the nuances...