Online payment fraud: from anomaly detection to risk management

Abstract Online banking fraud occurs whenever a criminal can seize accounts and transfer funds from an individual’s online bank account. Successfully preventing this requires the detection of as many fraudsters possible, without producing too false alarms. This is challenge for machine learning owing to extremely imbalanced data complexity fraud. In addition, classical methods must be extended, minimizing expected financial losses. Finally, only combated systematically economically if risks costs in payment channels are known. We define three models that overcome these challenges: learning-based detection, economic optimization results, risk model predict while considering countermeasures. The were tested utilizing real data. Our alone reduces unexpected losses aggregated by 15% compared benchmark consisting static if-then rules. Optimizing machine-learning further 52%. These results hold with low positive rate 0.4%. Thus, framework viable business perspective.