Securing The Neural Ledger: Deep Learning Approaches For Fraud Detection And Data Integrity In Tax Advisory Systems

Abstract

The pace of technological progress in business is leading to rapid operational changes, including the emergence of the connected audit. The primary concern that arises within the connected audit framework relates to the classification or integrity of data from a trusted perspective. Traditional data review in the audit process is time-consuming, expensive, and limited in its ability to detect misclassifications, manipulations, or distortions. This is because testing data validity, accuracy, and integrity requires a substantial amount of observational understanding to be performed effectively. Multiple methodologies that rely on various types of data analytics and machine learning models have become the focus of researchers due to the limitations of traditional methods and the massive amounts of data routinely provided by today's business organizations. Given the significant proportion of errors and inconsistencies within the data across tax firms and the impossibility of running an infinite number of tests on the tax form before it is submitted, the use of fraud detection and data integrity models can act as preliminary data review and validation tools for tax advisory systems. For these reasons, the present research endeavors to investigate the fraud detection and data integrity concept of DeepHybrid, applying deep learning models.