Graph Fraud

The core of how Equitus, Wallaroo, and IBM's Power11 with Spyre cards can create a powerful, end-to-end solution for a use case like fraud detection. This is where the synergy becomes particularly clear and valuable. Here’s a breakdown of how each component contributes:

1. Equitus KGNN: The Fraudster's Worst Nightmare

Traditional fraud detection often struggles because it looks at individual transactions in isolation. A single fraudulent transaction might look normal, but it's the connections—the web of relationships between accounts, devices, and people—that reveal the true picture. This is where Equitus KGNN shines.

• Data Unification and Context: Fraud data is notoriously siloed. Equitus KGNN automatically ingests raw, disconnected data from various sources (customer information, transaction logs, device IDs, account history, etc.) and transforms it into a unified knowledge graph. This graph represents all of the entities and their relationships.

• Discovering Hidden Connections: With this knowledge graph, the AI can see beyond a single transaction.¹ It can, for example:

  • Identify a seemingly legitimate transaction made by a customer who has a series of recent transactions with a newly linked, known fraudulent account.

  • Uncover a network of accounts sharing the same device ID, IP address, and common fraudulent spending patterns, even if each individual transaction looks normal.

  • Detect "mule" accounts that are used to move money around in a complex money-laundering scheme by analyzing the flow of funds across multiple entities.

• AI-Ready Data: The structured, context-rich data generated by KGNN is perfectly optimized for training and inference.² This eliminates the massive, time-consuming effort of manual data wrangling and feature engineering that is a major bottleneck in AI development.³

2. Wallaroo.ai: The Real-Time Fraud Fighting Command Center

Once the models are trained on the KGNN-generated data, you need a way to deploy and manage them in real-time, at scale. That’s Wallaroo.ai's role.

• High-Speed Inference: Fraud detection is a latency-sensitive use case. You need to make a decision in milliseconds to approve or deny a transaction. Wallaroo.ai's platform is built for high-performance, low-latency inference.⁴ It can handle a massive volume of real-time transactions and score them against the fraud model instantly.

• Simplified MLOps: Wallaroo.ai provides a central hub for managing the entire model lifecycle.⁵ Data scientists can quickly deploy new or updated fraud models without requiring extensive engineering support. This agility is crucial for keeping up with evolving fraud tactics.

• Monitoring and Governance: Wallaroo.ai continuously monitors the performance of the deployed fraud models.⁶ If a model's accuracy begins to degrade—a phenomenon known as model drift, which happens when fraudsters change their methods—Wallaroo can send alerts and enable the team to quickly retrain and redeploy an updated model.

3. IBM Power11 with Spyre Cards: The Secure, High-Performance Engine

The entire solution runs on the IBM Power11 server, which provides the secure and powerful infrastructure to make it all possible.

• Processing Power for KGNN: Equitus KGNN runs on the Power11 hardware, leveraging the processor's built-in Matrix Math Accelerator to efficiently process the complex relationships and connections within the knowledge graph.⁷ This allows for rapid data structuring without the need for additional accelerators.

• Accelerating Inference with Spyre: For the most complex, deep-learning models used in fraud detection, the optional Spyre accelerator cards become a game-changer.

  • Low-Latency, High-Throughput: Spyre cards are specifically designed for efficient AI inference. This allows the system to process a higher volume of transactions with even lower latency, which is critical for preventing fraud in real-time.

  • Power Efficiency: Spyre's low-wattage design provides a powerful and energy-efficient alternative to traditional GPUs, reducing operational costs.⁹

  • Data Security: The most important advantage is that all of this processing happens on-premise, in a highly secure environment. The Power11 server with Spyre cards allows financial institutions to keep sensitive transaction data secure and in compliance with regulations, eliminating the risks associated with moving data to a public cloud for processing.¹¹

The Complete Synergy

In this powerful ecosystem, Equitus KGNN provides the "brain" by turning disconnected data into actionable intelligence.¹² Wallaroo.ai provides the "nervous system" by managing the real-time AI and ensuring operational excellence. And IBM Power11 with Spyre cards provides the "muscle" to execute these AI tasks at scale, securely, and with a level of performance that is difficult to match in a traditional data center.¹³

By working together, these technologies enable IBM Power users to directly address a high-value business problem like fraud detection with a single, integrated, and highly optimized solution.