Conditional Adversarial Fragility in Financial Machine Learning under Macroeconomic Stress

TL;DR

This paper demonstrates that adversarial robustness in financial machine learning is not a static property but is conditioned on macroeconomic regimes. By introducing a regime-aware evaluation framework and a novel Risk Amplification Factor, it shows that adversarial perturbations cause nearly double the degradation under stress than under calm conditions, even when baseline performance is unchanged. The authors also add a governance layer with a Semantic Robustness Index that combines SHAP stability and LLM-based narrative consistency, highlighting explanation instability as an early warning signal. The work argues for regime-aware robustness testing and adaptive governance to better manage model risk in high-stakes, nonstationary financial environments.

Abstract

Machine learning models used in financial decision systems operate in nonstationary economic environments, yet adversarial robustness is typically evaluated under static assumptions. This work introduces Conditional Adversarial Fragility, a regime dependent phenomenon in which adversarial vulnerability is systematically amplified during periods of macroeconomic stress. We propose a regime aware evaluation framework for time indexed tabular financial classification tasks that conditions robustness assessment on external indicators of economic stress. Using volatility based regime segmentation as a proxy for macroeconomic conditions, we evaluate model behavior across calm and stress periods while holding model architecture, attack methodology, and evaluation protocols constant. Baseline predictive performance remains comparable across regimes, indicating that economic stress alone does not induce inherent performance degradation. Under adversarial perturbations, however, models operating during stress regimes exhibit substantially greater degradation across predictive accuracy, operational decision thresholds, and risk sensitive outcomes. We further demonstrate that this amplification propagates to increased false negative rates, elevating the risk of missed high risk cases during adverse conditions. To complement numerical robustness metrics, we introduce an interpretive governance layer based on semantic auditing of model explanations using large language models. Together, these results demonstrate that adversarial robustness in financial machine learning is a regime dependent property and motivate stress aware approaches to model risk assessment in high stakes financial deployments.

Figures (1)

• Figure 1: Conceptual architecture of regime-aware adversarial evaluation. Macroeconomic stress conditions model behavior through regime conditioning, while identical adversarial perturbations induce asymmetric performance degradation and explanation instability. Joint evaluation of predictive fragility and semantic drift yields conditional risk amplification metrics, which translate into economically material exposure.