Watch Out! Simple Horizontal Class Backdoors Can Trivially Evade Defenses

arXiv:2310.00542v2 Announce Type: replace
Abstract: All current backdoor attacks on deep learning (DL) models fall under the category of a vertical class backdoor (VCB) -- class-dependent. In VCB attacks, any sample from a class activates the implanted backdoor when the secret trigger is present. Existing defense strategies overwhelmingly focus on countering VCB attacks, especially those that are source-class-agnostic. This narrow focus neglects the potential threat of other simpler yet general backdoor types, leading to false security implications. This study introduces …

activates arxiv attacks backdoor backdoor attacks backdoors can class cs.cr cs.lg current deep learning defense defenses defense strategies evade focus sample secret simple strategies trigger under watch