This project focuses on detecting cheating behavior in online games using unsupervised machine learning–based anomaly detection techniques.

The system analyzes multi-dimensional, user-specific gameplay data including attributes such as age group, player rank, geographic location, progression patterns, and behavioral metrics—without relying on labeled examples of cheating.

By modeling normal player behavior across 7-10 feature dimensions, the algorithm identifies statistically abnormal patterns that may indicate unfair play, exploitation, or automated behavior. This approach enables scalable and adaptive fraud detection, particularly useful in dynamic gaming environments where cheating strategies evolve rapidly and labeled datasets are limited or unavailable.

The solution improves game integrity, enhances player trust, and supports data-driven moderation without manual rule-based enforcement.