A Secure Framework for Strengthening Fraud and Error Detection in IT Audit Systems in U.S. Banks
Natasha Mwanandimayi
*
Department of Accounting, College of Business, Michigan Technological University, Michigan, United States of America.
Isaiah Osemudiamen Okogun
Department of Mathematical Science, College of Science and Arts, Michigan Technological University, Michigan, United States of America.
Rutendo Talent Sithole
Department of Accounting, College of Business, Michigan Technological University, Michigan, United States of America.
Claudious Mufandaidza
Department of Accounting, College of Business, Michigan Technological University, Michigan, United States of America.
*Author to whom correspondence should be addressed.
Abstract
Background: The US banking sector is continually becoming more dependent on complex IT infrastructure, which exposes it to greater risks of sophisticated fraud and transactional errors. The heavy reliance on technology has rendered traditional audit mechanisms obsolete and unable to track evolving threats. Consequently, this failure puts both financial transaction integrity and regulatory compliance at risk.
Objective: The main goal of this paper is to create a complete framework to improve the detection of fraud and errors in IT audit systems. To achieve this, the limitations of current audit systems and real-world implementation challenges were examined. The proposed framework was then validated using actual data, while advanced analytics were explored to enhance detection accuracy.
Method: The study used a mixed-methods approach to cover the audit landscape in a comprehensive manner. The study combined a theoretical integration of standards like COBIT, COSO, and TAM with semi-structured interviews of industry professionals. Besides that, real-world and simulated data scenarios were utilised in the study to both apply and corroborate the framework, thus conducting a thorough testing of its effectiveness, accuracy, and efficiency in a realistic banking environment.
Result and Conclusion: Findings reveal that current IT audit systems are limited by their reliance on static, rule-based procedures, which often fail to detect complex fraud schemes due to poor data integration. However, the research demonstrates that deploying advanced analytics—specifically big data and machine learning—significantly expands the capability to process massive datasets and uncover concealed irregularities.
Results of the validation experiments showed that a significant factor in successful uptake was the reliance on very specific control components featuring stringent access restrictions, proper segregation of tasks and automated monitoring on a continual basis. The current research has confirmed that a transition to a continuous, data- driven audit framework gives the U.S. banking sector an opportunity to change its mode from being reactive in compliance to being proactive in risk management.
Keywords: IT audit systems, fraud detection, U.S. Banking sector, data analytics, machine learning, continuous monitoring, COSO, COBIT