Stolen credit cards are a major problem faced by many companies, including Uber. Karthik Ramasamy and Lenny Evans detail a new weapon against stolen credit cards that uses computer vision to scan credit cards, verifying possession of the physical card with basic fake card detection capabilities. There are few labeled datasets of credit cards, so Uber uses transfer learning on public datasets to generate synthetic data to train its models.
OCR is a well-studied problem, and there are many approaches to building an OCR framework. Karthik and Lenny discuss the benefits and trade-offs of various approaches as well as challenges they faced with their model, including running it on mobile devices, handling various lighting and background conditions, and making the model-size footprint small. If you want to incorporate object and text detection models running on-device into your platform, this is the talk for you.
Karthik Ramasamy leads a data science team at Uber focusing on solving fraud problems using machine learning. His team builds advanced machine learning models like semisupervised and deep learning models to detect account takeovers and stolen credit cards. Previously, Karthik was a cofounder of LogBase, where he worked on real-time analytics infrastructure and built models to rate drivers based on their driving behavior, and a founding member of the LinkedIn security team, where he developed various security products, with a particular focus on anti-automation efforts.
Lenny Evans is a data scientist at Uber focused on the applications of unsupervised methods and deep learning to fraud prevention, specifically developing anomaly detection models to prevent account takeovers and computer vision models for verifying possession of credit cards.
©2018, O'Reilly Media, Inc. • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • email@example.com