How to Use Technology to Prevent Fraud

Here's how Stripe detects frauds with a 99.9% accuracy in 100 milliseconds (that too by checking over 1000 parameters for one transaction) Fraud detection in online payments isn’t just about stopping bad transactions it’s about doing it fast, at scale, and without blocking legitimate users. Stripe’s fraud prevention system, Radar, evaluates 1,000+ signals within 100 milliseconds to make decisions. Here’s how it works and why it’s so effective: 1. ML Models That Learn and Scale Stripe started with simple ML models (logistic regression) but quickly scaled to hybrid architectures combining: –XGBoost for memorization (catching known patterns). –Deep Neural Networks (DNNs) for generalization (handling unseen patterns). –Key Problem: XGBoost couldn’t scale or integrate modern ML techniques like transfer learning and embeddings. –The Solution: Stripe moved to a multi-branch DNN-only architecture inspired by ResNeXt. This setup allowed it to memorize patterns while staying scalable. It reduced training times by 85%, enabling multiple experiments in a single day instead of overnight runs. 2. Learning From Real Fraud Patterns Radar doesn’t just rely on static rules, it learns from data across Stripe’s network. –Engineers analyze fraud attacks in detail, e.g., patterns of disposable emails or repeated card testing. –Features like IP clustering and velocity checks were added to detect suspicious activity. –Fraud insights are shared across the network, so lessons learned from one business protect others automatically. Example: Analyzing IP patterns helped detect high-volume attacks where fraudsters used multiple stolen cards from the same source. 3. Scaling With More Data, Not Just Smarter Models Stripe realized that more training data could unlock better performance, similar to modern LLMs like GPT models. It tested scaling datasets by 10x and 100x. Result? Performance kept improving, confirming that larger datasets and faster training cycles work better than complex rules alone. Key Insight: Bigger datasets help uncover rare fraud cases, even if they occur in only 0.1% of transactions. 4. Explaining Fraud Decisions Clearly Fraud systems often act like black boxes, leaving businesses guessing why a payment failed. Stripe built Risk Insights to provide clear explanations: –Shows features contributing to fraud scores like mismatched billing and shipping addresses. –Displays maps and transaction histories for visual context. –Enables custom rules to fine-tune fraud checks for specific business needs. Result: Businesses trust Radar’s decisions because they can see why a payment was flagged. 5. Constant Adaptation to Stay Ahead Fraud patterns evolve, so Stripe built Radar to adapt in real time: Uses transfer learning and multi-task learning to generalize better. Incorporates insights from the dark web and emerging fraud tactics. Continuously retrains models without disrupting performance.

There are too many fraud prevention guides out there. But not enough clear, step-by-step checklists. So, here’s a simple one for you to follow daily: 1. 𝗠𝗮𝗽 𝘆𝗼𝘂𝗿 𝘂𝘀𝗲𝗿 𝗷𝗼𝘂𝗿𝗻𝗲𝘆 Figure out where fraud might sneak in From first click to checkout. 2. 𝗦𝗽𝗼𝘁 𝘄𝗲𝗮𝗸𝗻𝗲𝘀𝘀𝗲𝘀 Identify where you’re most exposed. Are fake accounts an issue? Payment fraud? 3. 𝗦𝗲𝘁 𝘂𝗽 𝗿𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝗮𝗹𝗲𝗿𝘁𝘀 Catch fraud fast by monitoring unusual behavior Sudden changes or odd transactions. 4. 𝗟𝗮𝘆𝗲𝗿 𝘆𝗼𝘂𝗿 𝗽𝗿𝗼𝘁𝗲𝗰𝘁𝗶𝗼𝗻𝘀 Use a mix of tools. Like extra login security and behavior tracking. 5. 𝗥𝗲𝗴𝘂𝗹𝗮𝗿𝗹𝘆 𝗿𝗲𝘃𝗶𝗲𝘄 𝗮𝗻𝗱 𝘂𝗽𝗱𝗮𝘁𝗲 Fraudsters adapt quickly. Your strategy should, too. Keep it fresh. You don't need fluffy guides, you need actionable steps.

Fraudulent activities pose a significant threat to many businesses, making it crucial to detect and prevent them to protect both the company's reputation and bottom line. In a blog post by the engineering team from Booking.com, they share their innovative approach to combating fraud using graph technology. The rationale behind leveraging graph technology for fraud detection is straightforward: often, there are hidden links between various actors, identifiers, and transactions. For example, if an email address has been previously associated with fraudulent activity, it provides valuable context for future detection. This interconnected nature makes graph-based features highly effective for identifying fraud. The team at Booking built a graph using historical data, such as reservation requests. In this graph, nodes represent transaction identifiers like account numbers and credit card details, while edges connect identifiers that have been observed together before. When assessing fraud risk, they query the graph database to build a local graph centered around the request identifier, which helps to evaluate the likelihood of fraudulent behavior. One aspect that stands out is the dynamic visual representation of how the graph evolves with customer interactions, making it easier to understand the benefits of graph technology in fraud detection. It serves as a nice introduction to the potential of graph technology in combating fraudulent activities. #machinelearning #graph #datascience #analytics #fraud #detection – – –  Check out the "Snacks Weekly on Data Science" podcast and subscribe, where I explain in more detail the concepts discussed in this and future posts:    -- Spotify: https://lnkd.in/gKgaMvbh   -- Apple Podcast: https://lnkd.in/gj6aPBBY    -- Youtube: https://lnkd.in/gcwPeBmR https://lnkd.in/gQAwSz7D

Why do agencies need advanced data analytics tools to prevent fraud? Here’s a case example. The Postal Service IG uncovered a $2.3 million fraud scheme involving two brothers who submitted 18,000 Priority Mail insurance claims to the Postal Service. The scheme? They mailed low-value items or simply empty packages and reported them as damaged or lost. They also backed their claims with fictitious invoices and fraudulent images of the contents. The Postal Service, in good faith, paid up to $100 per claim plus shipping costs. If the Postal Service had used anomaly detection algorithms, open source intelligence and document authentication tools, they would have caught this scheme and prevented the losses. IGs play a vital role in ex post facto analysis but if we want to move from “pay and chase” to prevention, we need to arm agencies with the analytics tools they need. #fraudprevention #dataanalytics