Spatiotemporal attacks for embodied agents

Liu, A. and Huang, T. and Liu, X. and Xu, Y. and Ma, Y. and Chen, X. and Maybank, Stephen J. and Tao, D. (2020) Spatiotemporal attacks for embodied agents. Lecture Notes in Computer Science 12362 , pp. 122-138. ISSN 0302-9743.

Abstract

Adversarial attacks are valuable for providing insights into the blind spots of deep learning models and help improve their robustness. Existing work on adversarial attacks have mainly focused on static scenes; however, it remains unclear whether such attacks are effective against embodied agents, which could navigate and interact with a dynamic environment. In this work, we take the first step to study adversarial attacks for embodied agents. In particular, we generate spatiotemporal perturbations to form 3D adversarial examples, which exploit the interaction history in both the temporal and spatial dimensions. Regarding the temporal dimension, since agents make predictions based on historical observations, we develop a trajectory attention model to explore scene view contributions, which further help localize 3D objects appeared with the highest stimuli. By conciliating with clues from the temoral dimension, along the spatial dimension, we adversarially perturb the physical properties (e.g., texture and 3D shape) of the contextual objects that appeared in the most important scene views. Extensive experments on the EQA-v1 dataset for several emboded tasks in both the white-box and the black-box settings have been conducted, which demonstrate that our perturbations have strong attack and generalization abilities.

Metadata

Statistics

DownloadsShow export options

Export as XMLJSONCSV

Activity Overview

6 month trend

241Downloads

6 month trend

119Hits

Additional statistics are available via IRStats2.

Archive Staff Only (login required)

Edit/View Item