Abstract: Transferability and imperceptibility of adversarial examples are pivotal for assessing the efficacy of black-box attacks. While diffusion models have been employed to generate adversarial ...