Towards Privacy Preserving Financial Fraud Detection
Stephanie Abanilla, Moitrayee Chatterjee, Shuvalaxmi Dass · 2023
Federated Learning (FL) has gained prominence in fields where safeguarding data privacy is of utmost importance, presenting a valuable approach for the detection of credit card or financial fraud. The detection of fraud in credit card transactions, as well as other digitized financial activities, plays a pivotal role in upholding the integrity of financial transactions, safeguarding the assets of individuals and businesses, and contributing to a more secure and trustworthy financial environment. The field of machine learning and deep learning is continuously advancing, offering promising solutions for bolstering fraud prevention and minimizing financial losses. Nevertheless, in situations where data owners are hesitant to share their data due to privacy regulations, security concerns, or the sensitive nature of the information involved, a privacy-preserving machine learning technique like FL can significantly enhance fraud detection. FL achieves this by harnessing the collective strength of multiple institutions' data without compromising the privacy of individual users. The re-search presented in this paper leverages the advantages of FL for financial fraud detection. The proposed FL model outperformed a conventional neural network in terms of precision, accuracy, loss and recall when identifying instances of financial fraud.