Quantifying the robustness of XAI in financial fraud detection. A systematic verification framework for assessing the stability and fidelity of additive local explanations (SHAP/LIME) under controlled perturbations.