Versatile Backdoor Attack with Visible, Semantic, Sample-Specific, and Compatible Triggers

arXiv:2306.00816v3 Announce Type: replace
Abstract: Deep neural networks (DNNs) can be manipulated to exhibit specific behaviors when exposed to specific trigger patterns, without affecting their performance on benign samples, dubbed \textit{backdoor attack}. Currently, implementing backdoor attacks in physical scenarios still faces significant challenges. Physical attacks are labor-intensive and time-consuming, and the triggers are selected in a manual and heuristic way. Moreover, expanding digital attacks to physical scenarios faces many challenges due to their sensitivity to visual distortions and the absence …

abstract arxiv attacks backdoor challenges cs.cr cs.cv labor networks neural networks patterns performance sample samples semantic type