Credit card fraud is a prevalent financial crime thatcauses substantial losses to individuals and financial institutions.Traditional fraud detection methods often rely on rule-basedsystems or supervised machine learning algorithms, which maynot be effective in detecting novel or evolving fraudulent patterns.This project proposes a novel approach for credit card frauddetection using a generative adversarial network (GAN) togenerate synthetic fraudulent transactions, feature engineeringtechniques to extract relevant features from the transactiondata, and anomaly detection methods to identify fraudulenttransactions based on their deviation from normal patterns. Theproposed approach involves three main phases: data prepara-tion and feature engineering, GAN-based data balancing, andanomaly detection. Experimental results demonstrate that theproposed approach outperforms traditional fraud detection meth-ods, achieving higher accuracy, precision, recall, and F1-score inidentifying fraudulent transactions. Additionally, the approach ismore robust to changes in the underlying data distribution andcan effectively detect novel or evolving fraudulent patterns.