Concept Drift Adaption for Online Game Chargeback Detection

The flourishing online game market, driven by the rapid development of the Internet and hardware performance, has attracted the attention of criminals. Online game service providers have become prime targets, facing significant losses due to malicious chargebacks. Unfortunately, the current approach is reactive, as providers can only block affected game accounts after they have been attacked. Although there is existing research on detecting malicious chargebacks using machine learning, these criminals intentionally evade detection, exacerbating the concept drift of game records. This research aims to address not only the enhancement of malicious chargeback detection in online games but also the detection and prevention mechanisms for concept drift. In this paper, we propose an adaptive learning model for concept drift in online game top-up fraud, with the goal of preventing malicious chargebacks in top-ups before any losses occur.