Event Streaming-Based Real-Time Fraud Detection System - Global Leading Game Company K
Leading Game Company K, which previously processed large-scale game logs through daily batch jobs, implemented a streaming platform to collect global server logs in real time and instantly detect abnormal behavior. This unified their operational solutions, reduced personnel resources, and strengthened fraud prevention and service stability.
Challenge
Leading Game Company K, operating a large-scale MMORPG, relied on daily batch processing to detect fraudulent activity, making real-time threat response impossible. This approach created gaps in safeguarding the game economy and ecosystem and imposed high costs due to reliance on external analytics solutions. The company urgently required a real-time system to unify log collection, analysis, and response.
Solution
The company implemented an event streaming architecture using the Kafka-based Confluent Platform, enabling real-time ingestion of game server logs. Fraudulent patterns, such as abnormal clicks, movements, or transactions, were instantly detected through KSQL-based streaming analytics. Data was efficiently partitioned by topic, and the system triggered immediate notifications and mitigation measures upon detecting suspicious activity.
Results
- Proactive fraud detection and mitigation: Protected in-game economies through instantaneous response mechanisms
- Cost reduction by eliminating external analysis solutions: Optimized license and data processing costs with in-house solutions
- Enhanced stability by unifying security infrastructure: Reduced points of failure and increased management efficiency
- Improved trust in the game ecosystem: Maintained a fair game environment, enhancing customer satisfaction and loyalty