Rule-Based Fraud Transaction Detection
Implement a fraud detection system that evaluates transactions against a sequential list of text-based rules, where each transaction is represented as a dictionary of attributes such as merchant name and card brand. Rules must be processed in order, defaulting to accept, and short-circuiting when an accept decision is reached.
Asked at:
Stripe
Question Timeline
See when this question was last asked and where, including any notes left by other candidates.
Mid July, 2026
Stripe
Staff
The interview consisted of 4 modules, but module completion was not the target—you received a new module upon completing each one. The platform was HackerRank with an inbuilt AI model (Sonnet and GPT-4.5 were available). The task was to evaluate whether a transaction was fraudulent based on a set of rules. Each transaction came as a dictionary with keys like merchant_name and cardbrand. You were given a rules list (text format strings) that you needed to process in order. The evaluation logic followed these principles: Process rules sequentially If an earlier rule leads to an "accept" decision, stop processing remaining rules Default outcome: accept First module had simple rules one key and one value in the condition. Second module had a complex condition with only AND or OR as boolean.
Your account is free and you can post anonymously if you choose.