Photo of Feedzai's Catarina Godinha, demonstrating how Feedzai Railgun enhances fraud detection with large language models

With Railgun, Feedzai has changed the game for risk profiling and fraud detection. Solving the conundrum of the data storage hierarchy makes it possible for risk profiles, based on almost limitless data points, to be calculated and put into production without delay.

Learn how this patented streaming engine delivers a solution that addresses the challenges of accuracy, latency, and scalability at the same time.

What is Feedzai’s Railgun?

Streaming systems are designed to process data as it is ingested, eliminating the need to store all the data beforehand. This real-time processing capability enables organizations to make timely decisions and respond swiftly to emerging situations. Streaming systems are vital for numerous applications in diverse industries by analyzing data on the fly, from finance and eCommerce to transportation and cybersecurity.

Put it another way. If data is oil, streaming systems are the refinery that converts data into power. Feedzai’s Railgun is a revolutionary multi-patented risk engine capable of refining an abundance of data that fuels fraud detection and customer trust. By seamlessly integrating, processing, and extracting insights from real-time data, Railgun enables accurate fraud detection with unprecedented speed, safeguarding businesses from financial losses and protecting customers from financial harm.

Current Market Challenges with Machine Learning and Fraud Detection

Many traditional streaming systems were designed before many banks underwent their digital transformation. As such, these systems need help to keep up with the demands of real-time data processing, resulting in three critical problems.

Increasing Transaction Volumes and Costs

The market has witnessed a shift in focus from data collection to data processing. The democratization of banking and the adoption of new channels have significantly increased transaction volumes. This includes digital activity data like logins, clicks, and paths that generate vast amounts of data that are highly relevant for fraud detection context. Existing financial institutions experience constant growth, while new institutions cater to previously unmet needs. However, legacy systems must be adequately equipped to handle the massive influx of data, resulting in high costs and infrastructure limitations.

Rise of Fraud Threats

As transaction volumes grow, fraud threats become more sophisticated and harder to detect. Instant earnings and rapid money movement create challenges in blocking fraud because they require real-time actions and intelligent systems. Financial institutions need to be agile in responding promptly to mitigate emerging fraud threats.

Technological Market Standards

Off-the-shelf technologies often fail to scale effectively with growing transaction volumes or accurately detect fraud. Maintaining profitability and coping with infrastructure limitations force compromises on either scalability or accuracy. Financial institutions are left with the difficult choice of prioritizing one over the other.

Key Features of Feedzai’s Railgun

Railgun is a pioneering streaming system engineered to overcome the challenges of accuracy, latency, and scalability in mission-critical fraud prevention. 

Diagram explaining how Feedzai's Railgun reinvents fraud detection by both calculating accurate real-time metrics in both short and long periods.
Diagram explaining how Feedzai's Railgun reinvents fraud detection by both calculating accurate real-time metrics in both short and long periods.

Railgun sets a new standard for real-time data processing by implementing various innovative techniques. Some of the key features that make Railgun unique include:

  1. Real-Time Sliding Window Algorithm. Railgun employs a sophisticated real-time sliding window algorithm to process thousands of events per second. This algorithm enables Railgun to compute profiles on data streams within long real-time windows (up to months), continuously updating the analysis as new data arrives. This approach maximizes data accuracy while keeping latencies low, making Railgun an ideal solution for time-sensitive and mission-critical applications.
  2. Event Reservoir. Profiles must be updated in real-time within a long period. For this to happen, critical data must be available in fast memory (e.g., RAM) for easy access. The remaining data, stored in hard drives, is read ahead of its need using sophisticated caching mechanisms. With this approach, memory consumption is optimal, and we can support profiles with extremely long windows with the same performance and accuracy as small windows, ensuring total flexibility for fraud risk strategies. With this approach, memory consumption is optimized and can support profiles with extremely long windows with the same performance and accuracy as small windows, ensuring total flexibility for fraud risk strategies.
  3. Distributed Architecture & Fault-Tolerant Design. Leveraged by a distributed architecture – operating across multiple instances or nodes – Railgun handles vast amounts of data by distributing the processing workload efficiently, accommodating growing data streams seamlessly. It is also fault-tolerant. By employing redundancy and intelligent error handling, Railgun can continue operating with zero downtime in the face of hardware or software failures, providing uninterrupted data processing capabilities.

How Railgun Addresses the Data and Storage Struggle

Given their limitations, many legacy systems struggle to provide a holistic view of a customer’s profile. They often simplify data, sacrificing accuracy or latency. Real-time metrics are calculated with recent data, while access to historical data is limited. This limitation makes adjusting to rapidly evolving fraud threats complex because the systems can’t access all relevant information in real time.

Machine learning models need access to new and recent data to make accurate decisions. This necessitates using profiles that capture essential information, such as transaction history over specific periods, allowing for comparisons between current and typical behaviors. However, storing these profiles for numerous customers, cards, and terminals requires significant memory space. 

These storage issues force difficult compromises, such as storing limited information or reducing the time window for data storage. It’s a computer science phenomenon known as the “hierarchy of storage.” This means there’s storage that can be quickly accessed but is more expensive. Alternatively, there is storage that can be accessed slower but is more affordable.

A Solution for Storage Hierarchy

Railgun addresses the hierarchy of storage challenges by efficiently storing and accessing previous events. This enables the computation of accurate metrics regardless of the data window size. Financial institutions can rapidly update their risk strategies with this capability, reducing the time required from days to hours. Railgun’s ability to recompute only new or updated metrics ensures swift implementation of changes, bringing unparalleled speed and efficiency to the fraud detection landscape.

Financial institutions that implement Railgun stand to gain several key benefits, including:

  • Greater Agility. Railgun improves the speed of risk strategy publication by as much as three times, enabling swift response to emerging fraud threats.
  • Enhanced Accuracy. Real-time calculations disregarding the timeframe and availability of comprehensive data provide FIs with better observability and understanding of customer behaviors, facilitating the detection of suspicious activities.
  • Low Latency. Railgun ensures that rules become effective almost immediately and enables risk recalculation without imposing a heavy burden on the system.
  • Scalability: With Railgun’s low latency, FIs can confidently handle increasing transaction volumes without compromising accuracy or latency, achieving granular profiling.

Prioritizing Fraud Detection

With Railgun, Feedzai has positioned customer needs as a top priority in fraud detection. Railgun provides financial institutions with meaningful data to quickly uncover suspicious behaviors and combat financial crime. 

By leveraging Railgun, financial institutions gain the agility to respond rapidly to emerging fraud threats with the following capabilities:

  • Clear data-based insights. FIs gain a clearer picture of who the customer is and their behaviors due to more new data available in real-time.
  • Earlier fraud detection. Detect fraud faster and more accurately by comparing recent and past behaviors from earlier periods in real-time. 
  • Stop fraud faster. Tackle fraud in real-time with information updated to the millisecond.

Feedzai’s Railgun provides a unique balance of scalability, low latency, and accuracy. This makes Railgun the ideal choice for financial institutions operating in high-risk environments. Businesses gain the agility to respond rapidly to fraud threats, understand customer behaviors in real time, and significantly reduce fraud losses. Railgun saves time and money and raises the standard for fraud detection, enabling businesses to grow confidently while keeping pace with evolving fraud trends.