Towards Resilient Safety-driven Unlearning for Diffusion Models against Downstream Fine-tuning | Read Paper on Bytez