Synthetic Data Generation for Fraud Detection using GANs
This work addresses the challenge of insufficient high-quality data for fraud detection in the gambling industry, though it is incremental as it applies an existing method to a specific domain.
The paper tackles the problem of class imbalance in fraud detection for online gambling by proposing a GAN-based system to generate synthetic data, which outperforms density-based oversampling methods and improves classification performance on benchmark and real-world datasets.
Detecting money laundering in gambling is becoming increasingly challenging for the gambling industry as consumers migrate to online channels. Whilst increasingly stringent regulations have been applied over the years to prevent money laundering in gambling, despite this, online gambling is still a channel for criminals to spend proceeds from crime. Complementing online gambling's growth more concerns are raised to its effects compared with gambling in traditional, physical formats, as it might introduce higher levels of problem gambling or fraudulent behaviour due to its nature of immediate interaction with online gambling experience. However, in most cases the main issue when organisations try to tackle those areas is the absence of high quality data. Since fraud detection related issues face the significant problem of the class imbalance, in this paper we propose a novel system based on Generative Adversarial Networks (GANs) for generating synthetic data in order to train a supervised classifier. Our framework Synthetic Data Generation GAN (SDG-GAN), manages to outperformed density based over-sampling methods and improve the classification performance of benchmarks datasets and the real world gambling fraud dataset.