Advancing Cheque Fraud Detection and Localization: A Holistic Approach Integrating CNN and Grad-CAM

The primary objective of this research is to enhance fraud detection in cheques by implementing a comprehensive approach that addresses anomalies in formatting, signatures, and other critical elements. With a focus on fortifying bank applications against financial malpractices, our goal is to develop a robust fraud localization system. To achieve this, we will utilize a CNN implemented by Verafin for efficient fraud localization cheques. We aim to approximate and localize areas within the cheque that may be susceptible to fraud, leveraging techniques such as the Grad-CAM method. This involves analyzing important regions in cheque images to identify potential fraudulent elements, such as forged signatures or irregular formatting. By combining classification with detailed localization, our research aims to provide a comprehensive solution for enhancing the integrity and security of cheque processing systems.

Faculty Supervisor:

Terrence Tricco

Student:

Partner:

NASDAQ Canada Inc

Discipline:

Computer science

Sector:

Artificial Intelligence

University:

Memorial University of Newfoundland

Program: