Fraud is an incredibly difficult problem for payment processors, for both technical and business reasons. It's well worth studying if you're into this sort of t

Fraud is an incredibly difficult problem for payment processors, for both technical and business reasons. It's well worth studying if you're into this sort of thing. Fraud is a sensitivity/specificity problem. For sensitivity, Stripe wants to detect as many actual fraud cases as possible, because undetected fraud is expensive. For specificity, they want as few false positives as possible, because false positives mean angry customers who feel mistreated. Generally, there's a sensitivity/specificity tradeoff, and priorities need to be set (which is worse - letting criminals get away, or frustrating honest customers?). But at such a scale, and with so many creative minds working on new mechanisms for fraud, it's impossible to achieve perfect numbers on either side.