Z. Yan, “DeHiB: Deep Hidden Backdoor Attack on Semi-supervised Learning via Adversarial Perturbation”, AAAI, vol. 35, no. 12, pp. 10585-10593, May 2021.