TAT: Targeted backdoor attacks against visual object tracking

Ziyi Cheng, Baoyuan Wu, Zhenya Zhang, Jianjun Zhao

Abstract: Highlights•We propose a novel and effective backdoor attack approach, named TAT, against visual object tracking.•We design a basic version of TAT in order to satisfy the effectiveness and stealthiness properties of backdoor attacks against VOT tasks.•We design a defense-aware version of TAT that can counter two representative backdoor defense techniques.•We perform large-scale experiments, and comprehensively evaluate the effectiveness and stealthiness of TAT.