On electronic game platforms, different payment transactions have differentlevels of risk. Risk is generally higher for digital goods in e-commerce.However, it differs based on product and its popularity, the offer type(packaged game, virtual currency to a game or subscription service), storefrontand geography. Existing fraud policies and models make decisions independentlyfor each transaction based on transaction attributes, payment velocities, usercharacteristics, and other relevant information. However, suspicioustransactions may still evade detection and hence we propose a broad learningapproach leveraging a graph based perspective to uncover relationships amongsuspicious transactions, i.e., inter-transaction dependency. Our focus is todetect suspicious transactions by capturing common fraudulent behaviors thatwould not be considered suspicious when being considered in isolation. In thispaper, we present HitFraud that leverages heterogeneous information networksfor collective fraud detection by exploring correlated and fast evolvingfraudulent behaviors. First, a heterogeneous information network is designed tolink entities of interest in the transaction database via different semantics.Then, graph based features are efficiently discovered from the networkexploiting the concept of meta-paths, and decisions on frauds are madecollectively on test instances. Experiments on real-world payment transactiondata from Electronic Arts demonstrate that the prediction performance iseffectively boosted by HitFraud with fast convergence where the computation ofmeta-path based features is largely optimized. Notably, recall can be improvedup to 7.93% and F-score 4.62% compared to baselines.
展开▼
