Machine unlearning as a security problem: Threats, defenses, and privacy-security-utility trilemma

Computer Science …, 2027 - Elsevier. … The survey further extends this analysis to large language models, federated learning, graph learning, reinforcement learning, and diffusion models, which do not …