Machine learning techniques, and in particular deep neural networks, created advances across a diverse range of application domains. Intriguing discoveries in deep learning research revealed that a variety of machine learning models, even simple regression models, are vulnerable and exhibit ‘intrinsic blind spots’ ‘ Adversarial attacks’ are deliberately designed to exploit such vulnerabilities and cause a machine learning model to make a mistake.
With the advent of deep adversarial learning, it became broadly accessible within reach of almost any individual with a computer. The early detection of such deepfakes is of high relevance in the context of societal disinformation and are of serious concern in democratic discourses. This holds in particular for the creation of adversarial journal entries to cover-up fraudulent2activities that might remain undetected by state-of-the-art ‘Computer Assisted Audit Techniques’.
Are financial audits vulnerable to adversarial attacks? And if so, are state-of-the-art CAATs able to detect such attacks? In this paper, we present a deep learning-based adversarial attack against CAATs. We regard this work to be an initial step towards the investigation of such future challenges in financial audits.
Auto108