Explore the ethical considerations and challenges in developing anti-fraud systems for mobile payments in this 15-minute IEEE conference talk. Delve into the complexities of payment fraud, algorithmic bias, and user-facing challenges in fraud detection. Examine the goals of a measurement study, key metrics, and the importance of end-user data in fraud evaluation. Learn about machine learning model design principles and system design considerations for effective fraud prevention. Analyze the results of an end-to-end fraud evaluation and the correlation between fraud decisions and frame rates. Gain insights into the limitations of algorithmic fairness and the need for innovative approaches in combating mobile payment fraud while maintaining ethical standards.
Doing Good by Fighting Fraud - Ethical Anti-Fraud Systems for Mobile Payments
Overview
Syllabus
Intro
Payment Fraud
Apps take fraud seriously
Fraud and algorithmic bias
User-facing challenges to overcome algorithmic bias
Are user-facing challenges the silver bullet?
Daredevil to the rescue!!!
Outline
Goals of the Measurement Study
Measurement platform
Key Metrics
Frame rate vs Success rate
Need the data from the end-user
Algorithmic Fairness doesn't work here
ML model design - Key principles
System design - Key principles
Results: End-to-end fraud evaluation.
Results: Is fraud decision correlated with Frame rate?
Conclusions
Taught by
IEEE Symposium on Security and Privacy
