THE DATA YOU NEED TO DETECT FRAUD IS THE DATA YOU ALREADY HAVE

Summary from the TLDR Product newsletter

You're already collecting behavioral signals to measure product activity: completion times, hesitation, copy/paste usage, autofill... Can you use the same data to stop bad actors? This guide by Persona [9] includes a 50-signal checklist, case studies of weak signals detecting sophisticated attacks, and frameworks for building fraud