This fork of MemGuard focuses on analyzing AI Blue Teaming techniques from a Red Teaming perspective. The goal is to explore how adversarial examples on confidence scores can mitigate privacy inference attacks and test its resilience against modern adaptive attacks.