Researchers tested attacks on cybersecurity classifiers and found that strong prediction accuracy can hide unstable SHAP explanations, especially in XGBoost models under black-box attacks.


arXiv.org
Beyond Gradient-Based Attacks: Adversarial Robustness and Explainability Stability in Cybersecurity Classifiers
Adversarial attacks on cybersecurity classifiers pose a dual threat: degrading predictions and destabilising the SHAP-based explanations that secur...












