PhD Position exploring manipulation-based defenses for deployed AI models. Study audit manipulation, randomized defenses, and defenses against model extraction, adversarial attacks, and reconnaissance. Develop theory and practical defenses with regulator-aware auditing; 36-month program.