Adversarial Robustness in Financial Machine Learning: Defenses, Economic Impact, and Governance Evidence

TL;DR

This paper assesses how small adversarial attacks affect financial machine learning models, revealing significant performance drops and exploring defenses like adversarial training to mitigate impact.

Contribution

It provides the first comprehensive evaluation of adversarial robustness in financial ML models, including empirical analysis and defense strategies.

Findings

Models suffer notable performance degradation under attacks.

Adversarial training partially recovers model performance.

Adversarial attacks impact discrimination and calibration metrics.

Abstract

We evaluate adversarial robustness in tabular machine learning models used in financial decision making. Using credit scoring and fraud detection data, we apply gradient based attacks and measure impacts on discrimination, calibration, and financial risk metrics. Results show notable performance degradation under small perturbations and partial recovery through adversarial training.