Smart Banking Audit Integrating Financial Entropy and Generative AI

Smart Banking Audit Integrating Financial Entropy and Generative AI

Ameneh.Raeiszadeh1 sayyed mohammadreza davoodi2

1) Department of Management, Deh.C, Islamic Azad University, Isfahan, Iran.
2) Department of Management, Deh.C, Islamic Azad University, Isfahan, Iran.

IOI Code : AAHD-ACZZCH
Publication : National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
    This study aims to introduce a novel framework for smart banking audit, termed “Physics-Informed Auditing.” Given the rapid acceleration of digital transformation, the proliferation of open banking, and the inherent complexity of real-time financial transactions, traditional audit methods—relying on cross-sectional sampling and historical financial statements—no longer meet contemporary regulatory requirements. Drawing inspiration from statistical physics and thermodynamics, this research employs Shannon information entropy as a metric to quantify disorder and friction within liquidity flows. The methodology is based on computational modeling and data-driven simulation, structured into four stages: representing the banking network as a directed graph incorporating Kirchhoff’s conservation laws; measuring entropy dynamics within sliding time windows; designing a Physics-Informed Generative Adversarial Network (PI-GAN) by injecting conservation constraints into the objective function; and empirical validation using standard datasets. Empirical results demonstrate that the proposed framework achieves an F1-score of 0.94, a false-positive rate reduction to 1.2%, and a response time of 18 milliseconds, significantly outperforming conventional models, including Logistic Regression, Random Forest, Deep Neural Networks, and standard GANs. Furthermore, in scenarios involving liquidity crises and tiered money laundering attacks, the model accurately simulated dynamic behaviors and predicted high-risk nodes with 92% precision. By enhancing interpretability and enabling real-time auditing, this framework represents a pivotal step in transitioning from reactive monitoring to predictive and resilient systems for the future banking ecosystem.

Keywords : physics informed auditing shannon information entropy physics informed generative adversarial networks (pi gans) flow conservation laws real time financial monitoring