Preprint
Aug 2026
Crushing the Evidence: A Dual-Penalty Evasion Framework for Fooling White-Box Explainable AI Auditors
A critical vulnerability is demonstrated by introducing a more potent white-box, gradient-regularized evasion attack framework that systematically crushes target feature attribution to near-zero, maintains>90% Attack Success Rates, and fundamentally bypasses Conditional Anomaly Detection.
Niraj Kumar, Harsh Kasyap
· 0 citations