Customer Identity Protection with PayPal’s Keyu Chen
Many of us have had the experience of someone hacking into our online purchasing system. It’s frustrating at the least and devastating at its worst. Companies like PayPal understand that this not only threatens the financial well-being of their client but also their own. Trust equals purchasing power and frequency. To combat this very modern problem, PayPal made use of a very modern problem solver in the form of Keyu Chen. A machine-learning scientist who has proven herself one of the most talented of her generation, Keyu is utilizing state-of-the-art software to outthink and outmaneuver would be identity theft fraudsters. Account takeover fraud (ATO) took a frightening one-hundred thirty-one percent upswing from 2021 to 2022; confirming that this online crime looks to expand rapidly unless obstructed. Keyu Chen’s development of a system known as Customer Identity Protection (CIP) model leverages a variety of data variables to stifle and rebuke would-be imposters. This development is a massive step forward in online security but don’t take our word for it, read on to have a clearer understanding of this game changing development.
Spotting a camouflaged threat was the goal Keyu intended for her CIP model, but what’s the ideal way to do this? This begins with data collection spanning from basic details like name, address, and payment methods to the IP addresses one typically visits, login history (where and when), and typical purchasing habits. This establishes a pattern of habits and a baseline identity against which unusual activity stands out. After the model learns what’s normal, its real job begins—spotting unusual behavior. The CIP Model is like a security guard who knows every regular customer and their individual habits. If someone behaves strangely, the model detects it. This could mean large unusual purchases, multiple password reset attempts, or other atypical behavior. The model then assesses a risk score to decide if it thinks a user has been hacked and if warranted, and asks the user to verify the activity.
What are the “nuts and bolts” of how this is achieved? Yes, it’s coding but much much more than this. After studying and analyzing millions of transactions to comprehend the process of fraudulent patterns, Keyu turned these red flags into variables that the CIP model could understand. For example, she wrote code to create variables like “typical purchase size” or “frequent login locations.” This is known as feature engineering, where raw data is turned into meaningful pieces of information. Keyu also wrote code in SQL and Python to “clean up” the data. This involved checking for missing data, making sure the numbers made sense, and ensuring the data was ready for model training. Using coding libraries such as scikit-learn or TensorFlow (tools for building machine learning models), the CIP model was then trained with machine learning algorithms. Finally, a system of tests and improvements sculpted the model into its final efficient and highly accurate form. The PayPal engineering team deployed the CIP model into the actual PayPal system where it now operates in real-time to check every login and transaction.
There’s a lot of humanity in the desire to create something that works with cutting-edge technology for the betterment of people. Keyu Chen is at the forefront of these advancements. Commerce has evolved at an almost unfathomable pace in recent years and we need safeguards in place like the CIP model to protect the average consumer. While technology advances, we need professionals who are able to think about how we can use it for security; at least as long as there are those who intend on using tech for malicious intent.
Writer : Clavin Hooney
