Privacy-aware adversarial defense with explainable AI for adversarial robustness in AI model
The Attention Concentration Score (ACS) is introduced, which measures how DP training shifts Transformer attention away from task-critical features toward non-functional ones, providing a mechanistic explanation of the trade-off.
Bilal Sardar, Shareeful Islam, Stefano Silvestri et al.
· Journal of Reliable Intellig... · 0 citations