JPMorganChase: Real-time fraud screening at massive scale (FSI315)
Key Takeaways
Payment Threat Landscape and Fraud Statistics
The payment threat landscape is constantly evolving as technology advances and cybercriminals become more sophisticated.
Key fraud trends include phishing/social engineering, card-not-present fraud, malware, and account takeover.
Recent fraud statistics show an 80% increase in payment fraud, 10% increase in commercial card fraud, and 71% of companies reporting being victims of email fraud.
How JPMorgan Chase Addresses the Problem
JPMorgan Chase's Trust and Safety Solutions protect clients throughout the commerce and payment lifecycle.
Key solutions include Validation Services, Confidence Score, Payment Control Center, and the Fraud Intelligence Platform.
The Fraud Intelligence Platform is a real-time fraud monitoring platform that leverages in-house rules, machine learning models, anomaly detection, and exception lists.
Building the Fraud Prediction Platform on AWS
Key requirements include real-time ingestion of payment data, analyzing data to detect fraud, autoscaling to handle high volumes, low latency, and 24/7 availability.
The architecture uses microservices for request ingestion, prioritization, data enrichment, detection, inference, and action handling.
Features are defined in a centralized Feature Registry and calculated using both batch and real-time processing.
The Model Serving API provides scores from deployed ML models.
Infrastructure Setup and Configuration
Leveraged AWS Load Balancer Controller to directly connect EKS pods to the Application Load Balancer, eliminating a reverse proxy layer.
Extensively used VPC Endpoints for low-latency access to other AWS services.
Used DynamoDB for large data lookups due to its scalable and low-latency performance.
Scaled EKS pods proactively based on CPU/memory and queue depth, rather than reacting to spikes.
Implemented multi-layer resiliency, including in underlying services, across Availability Zones, and across AWS Regions.
Cost Optimization
Treated cost as a non-functional requirement in all design decisions.
Implemented on-demand usage, with automated shutdown of unused environments.
Established observability metrics and a dedicated "Cost Champion" role to continuously optimize costs.
These cookies are used to collect information about how you interact with this website and allow us to remember you. We use this information to improve and customize your browsing experience, as well as for analytics.
If you decline, your information won’t be tracked when you visit this website. A single cookie will be used in your browser to remember your preference.