Abstract:Addressing the vulnerability of deep neural networks (DNNs) has attracted significant attention in recent years. While recent studies on adversarial attack and defense mainly reside in a single image, few efforts have been made to perform temporal attacks against video sequences. As the temporal consistency between frames is not considered, existing adversarial attack approaches designed for static images do not perform well for deep object tracking. In this work, we generate adversarial examples on top of video sequences to improve the tracking robustness against adversarial attacks under white-box and black-box settings. To this end, we consider motion signals when generating lightweight perturbations over the estimated tracking results frame-by-frame. For the white-box attack, we generate temporal perturbations via known trackers to degrade significantly the tracking performance. We transfer the generated perturbations into unknown targeted trackers for the black-box attack to achieve transferring attacks. Furthermore, we train universal adversarial perturbations and directly add them into all frames of videos, improving the attack effectiveness with minor computational costs. On the other hand, we sequentially learn to estimate and remove the perturbations from input sequences to restore the tracking performance. We apply the proposed adversarial attack and defense approaches to state-of-the-art tracking algorithms. Extensive evaluations on large-scale benchmark datasets, including OTB, VOT, UAV123, and LaSOT, demonstrate that our attack method degrades the tracking performance significantly with favorable transferability to other backbones and trackers. Notably, the proposed defense method restores the original tracking performance to some extent and achieves additional performance gains when not under adversarial attacks.

Diminishing-feature Attack: the Adversarial Infiltration on Visual Tracking

DIMBA: Discretely Masked Black-Box Attack in Single Object Tracking

F&F Attack: Adversarial Attack Against Multiple Object Trackers by Inducing False Negatives and False Positives

Pluggable Attack for Visual Object Tracking

Cooling-Shrinking Attack: Blinding the Tracker with Imperceptible Noises

Hijacking Tracker: A Powerful Adversarial Attack on Visual Tracking.

Efficient Adversarial Attacks for Visual Object Tracking

SPARK: Spatial-Aware Online Incremental Attack Against Visual Tracking

Robust Deep Object Tracking against Adversarial Attacks

Only Once Attack: Fooling the Tracker with Adversarial Template

One-Shot Adversarial Attacks on Visual Tracking With Dual Attention

Towards Universal Physical Attacks on Single Object Tracking

Blinding and blurring the multi-object tracker with adversarial perturbations

Ad2Attack: Adaptive Adversarial Attack on Real-Time UAV Tracking

Robust Tracking against Adversarial Attacks

Adversarial Bounding Boxes Generation (ABBG) Attack against Visual Object Trackers

Enhancing Tracking Robustness with Auxiliary Adversarial Defense Networks

Improved Real-Time Visual Tracking Via Adversarial Learning.

Evaluating the Robustness of LiDAR Point Cloud Tracking Against Adversarial Attack

LRR: Language-Driven Resamplable Continuous Representation against Adversarial Tracking Attacks

Multi-Model UNet: An Adversarial Defense Mechanism for Robust Visual Tracking