How NVIDIA’s new AI workflow is combating credit card fraud

AI-powered fraud detection in financial services: NVIDIA has launched a new AI workflow for detecting credit card transaction fraud, running on Amazon Web Services (AWS) and powered by the NVIDIA AI platform.

• The workflow aims to combat the growing problem of credit card fraud, which is expected to cause $43 billion in financial losses worldwide by 2026.
• By leveraging accelerated data processing and advanced algorithms, the new system can identify subtle patterns and anomalies in transaction data based on user behavior.
• The workflow is designed to improve accuracy and reduce false positives compared to traditional fraud detection methods.

Key features and capabilities: The NVIDIA AI workflow for fraud detection offers several advantages over traditional approaches, utilizing cutting-edge technologies to enhance performance and efficiency.

• It accelerates data processing, model training, and inference, integrating these components into a single, user-friendly software offering.
• The system uses NVIDIA RAPIDS Accelerator for Apache Spark, which can help payment companies reduce data processing times and costs.
• The workflow combines gradient-boosted decision trees (using XGBoost) with graph neural network (GNN) embeddings to improve fraud detection accuracy and reduce false positives.

Technical components: NVIDIA’s fraud detection workflow incorporates several proprietary technologies to deliver optimal performance and security.

• NVIDIA Morpheus Runtime Core library securely inspects and classifies incoming data, tagging patterns and flagging suspicious activity.
• NVIDIA Triton Inference Server optimizes the deployment of AI models in production, balancing throughput, latency, and utilization.
• These components, along with NVIDIA RAPIDS, are available through the NVIDIA AI Enterprise software platform.

Industry adoption and impact: Major financial institutions are increasingly turning to AI-powered solutions to combat fraud and protect their customers.

• American Express has been using AI for fraud detection since 2010, employing algorithms that monitor all customer transactions globally in real-time.
• European digital bank bunq utilizes generative AI and large language models for fraud and money laundering detection, achieving significantly faster model training speeds with NVIDIA accelerated computing.
• BNY Mellon recently deployed an NVIDIA DGX SuperPOD with DGX H100 systems to support fraud detection and other use cases.

Broader applications: While currently optimized for credit card transaction fraud, the NVIDIA workflow has potential for wider use in the financial services industry.

• The system could be adapted for detecting new account fraud, account takeover attempts, and money laundering activities.
• Systems integrators, software vendors, and cloud service providers can integrate the NVIDIA AI workflow to enhance their financial services applications and improve customer protection.

Future outlook and implications: The introduction of NVIDIA’s AI-powered fraud detection workflow signals a significant step forward in the financial industry’s fight against fraud.

• As AI models continue to grow in size and complexity, the need for cost-effective and energy-efficient computing power becomes increasingly crucial for organizations across industries, including financial services.
• The adoption of such advanced AI solutions by major financial institutions suggests a trend towards more sophisticated, real-time fraud detection systems that can keep pace with evolving threats in the digital financial landscape.