How to Craft Backdoors with Unlabeled Data Alone?

arXiv:2404.06694v1 Announce Type: cross
Abstract: Relying only on unlabeled data, Self-supervised learning (SSL) can learn rich features in an economical and scalable way. As the drive-horse for building foundation models, SSL has received a lot of attention recently with wide applications, which also raises security concerns where backdoor attack is a major type of threat: if the released dataset is maliciously poisoned, backdoored SSL models can behave badly when triggers are injected to test samples. The goal of this work …

arxiv backdoors cs.ai cs.cr cs.lg data