FiMMIA: Advancing Membership Inference in Multimodal AI Systems

This research explores membership inference attacks, a critical area for AI privacy. The study's focus on semantic perturbation across modalities suggests a sophisticated approach to uncovering vulnerabilities.