Credit Card Security with Generative Adversarial Networks: A Personalized Fraud Detection Framework

Banks and online retailers are being forced to use computerized fraud detection systems that mine vast transaction histories due to the recent spike in credit card-based online payment card scams. Machine learning (ML), which uses supervised binary classification algorithms that have been appropriately trained on pre-screened sample datasets in order to discriminate between fraudulent and legitimate events, appears to be one of the most effective means of identifying suspicious transactions. Generative Adversarial Networks (GANs) for content tailored generation in detecting credit card fraud (CCF) is discussed in this study. It compares the detection of CCF with and without employing GAN on four ML classifiers: Logistic Regression (LR), Random Forest (RF), XGBoost (XGB), and Naïve Bayes (NB). The results reveal that GAN with XGBoost performs better with accuracy at 98.6 %. On a general basis, GAN-created data improves the ability of ML models to identify fraud more accurately, minimizing cases of fraud not caught and improving system dependability.