Selective access control for medical image classification via friend-safe semantic-aware reversible adversarial examples
A unified framework called Friend-Safe Semantic-Aware Reversible Adversarial Examples (FS-SA-RAE), which integrates three complementary mechanisms: a friend-safe attack that employs dual-objective optimization to fool unauthorized (foe) models while preserving classification accuracy for authorized (friend) models, enabling selective access control and reversible data hiding based on histogram shifting.