Fraud Detection in Banking Applications : Machine Learning Approach

Abstract

Fintech, a burgeoning sector in the global market, increasingly relies on IT to automate financial services, enabling around-the-clock, easily accessible transactions for a global customer base. This growing trend highlights the pressing need for security, as these information-rich transactions are prime targets for fraud, causing significant harm to both consumers and service providers. In addressing these security concerns, Machine Learning (ML) methods are employed to detect anomalies and predict fraudulent activities in Fintech platforms. Our contribution to this field
includes a thorough evaluation of various anomaly detection techniques, analyzed across multiple real and synthetic datasets related to fraud. Our findings verify the variable success rates of ML in fraud detection, which led to a deeper investigation of the impact of specific features on the performance of these methods. Additionally, the paper analyzes the broader implications of these findings for the future security of Fintech services, acknowledging the high stakes of vulnerabilities that can lead to considerable financial and reputational losses. By proposing a machine learning
approach, we aim to advance fraud detection and develop strategies to mitigate and recover from such fraudulent incidents. The core of our research involved examining a plethora of intelligent algorithms on a publicly available dataset, which was adjusted to reduce class imbalances, thereby enhancing the precision of our proposed algorithm. The results of this comprehensive analysis may accelerate verification processes and reduce the incidence and impact of fraud in the banking sector.