NSUS Group Enhances Financial Fraud Prevention System with Amazon SageMaker

NSUS Group, a key player in the poker industry, operates a platform serving millions worldwide. Initially, it relied on Internet Data Centers (IDCs) for hosting, which met its needs but lacked scalability and efficiency. As the platform grew, NSUS Group saw an opportunity to strengthen its infrastructure, ensuring a seamless and secure gaming experience.

Compounding these challenges were the growing complexities of the business. NSUS needed its infrastructure to go beyond gaming to support advanced capabilities, such as user behavior analysis, and regulatory compliance. Financial fraud prevention became a critical priority as the company handled millions of daily financial transactions.

At the time, NSUS relied on manual investigative methods for fraud prevention. Skilled personnel analyzed historical data and user behavior logs, but this approach was resource-intensive, costly, and limited in scalability. 

Faced with these challenges, NSUS recognized the need for a more efficient and scalable solution. The company prioritized adopting automated ML pipelines to enhance fraud prevention capabilities, streamline operations, and reduce reliance on manual processes. However, transitioning to ML technologies was hindered by its legacy infrastructure and the high costs associated with implementing advanced systems.

Daniel Lim, chief product officer of NSUS Group, explained, “Previously, we relied heavily on human resources for our fraud prevention network. It became essential to adopt advanced technologies to scale our defense system and achieve greater efficiency.”

By addressing these challenges and focusing on innovation, NSUS set out to transform its operations, ensuring an even safer and more secure environment for its growing player base.

To address inefficiencies in its traditional anti-money laundering system, NSUS transitioned to a ML powered approach, building a predictive financial transaction management system. This system aimed to strengthen fraud prevention capabilities while streamlining operations. NSUS utilized Amazon Simple Storage Service (Amazon S3) to create a scalable, secure data lake for storing extensive user behavior data. To manage the machine learning (ML) workflow, NSUS adopted Amazon SageMaker for seamless data preprocessing, feature engineering, model training, hyperparameter optimization, and deployment.

During development, NSUS employed XGBoost-based training, a model widely used in e-commerce fraud prevention, and refined data with Apache Iceberg on AWS. By collaborating with AWS Partner Snowflake and adopting AWS Glue, NSUS streamlined data integration. 

Amazon SageMaker played a key role in optimizing hyperparameters, comparing models, tracking training history, and enabling collaborative model selection. Its testbed facilitated experimentation with diverse models and datasets, eliminating the need for specialized MLOps expertise. Additionally, NSUS used Amazon SageMaker Pipelines to manage inference workflows, storing models in Amazon S3 and implementing API-based real-time batch inference applications. 

By adopting Amazon SageMaker, NSUS transformed its labor-intensive, rule-based system into an automated ML-powered solution. This advanced system analyzes large datasets of user behavior in real-time, enabling proactive risk assessment and fraud prevention while significantly improving operational efficiency.

Lim says, “Amazon SageMaker allows NSUS to evaluate money laundering risks more effectively, building a stronger foundation for fraud prevention and fostering a healthier gaming environment.” The transition streamlined NSUS’s anti-money laundering operations and established a sophisticated fraud prevention system, ensuring a secure gaming experience for its global user base and demonstrating its commitment to innovation and regulatory compliance in the online poker industry.

NSUS successfully developed a ML powered financial fraud prevention system within three months using Amazon SageMaker. In the first month, the organization created an experimental environment to test and refine XGBoost-based models. Over the next two months, the team finalized the model, designed feature management processes, and built a production-grade data processing workflow and architecture.

Using Amazon SageMaker, NSUS rapidly prototyped and deployed its automated system, enabling efficient data insights and ML model development with minimal reliance on human resources. Amazon SageMaker’s Autopilot feature streamlined feature importance analysis, reducing the number of features from 270 to 80 and optimizing model performance. These advancements automated the fraud prevention system, improving business efficiency by 30 percent and cutting internal management costs, including human resource expenses, by over 80 percent. The system powered by Amazon SageMaker transformed NSUS’s manual processes into a preventive, data-driven fraud prevention framework. It achieved an Area Under Curve (AUC) score of 95 percent, highlighting its reliability.

Lim says, “As the reliability of our machine-learning models continues to improve, we will maximize business efficiency and enhance user experience by creating a seamless gaming environment. This financial fraud prevention system offers immense potential not only for our poker operations but also for expansion into the Multiplayer RNG Game market.”