Privacy Backdoors: Enhancing Membership Inference through Poisoning Pre-trained Models

Open in new window