English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

In autonomous driving, behavior prediction is fundamental for safe motionplanning, hence the security and robustness of prediction models againstadversarial attacks are of paramount importance. We propose a novel adversarialbackdoor attack against trajectory prediction models as a means of studyingtheir potential vulnerabilities. Our attack affects the victim at training timevia naturalistic, hence stealthy, poisoned samples crafted using a noveltwo-step approach. First, the triggers are crafted by perturbing the trajectoryof attacking vehicle and then disguised by transforming the scene using abi-level optimization technique. The proposed attack does not depend on aparticular model architecture and operates in a black-box manner, thus can beeffective without any knowledge of the victim model. We conduct extensiveempirical studies using state-of-the-art prediction models on two benchmarkdatasets using metrics customized for trajectory prediction. We show that theproposed attack is highly effective, as it can significantly hinder theperformance of prediction models, unnoticeable by the victims, and efficient asit forces the victim to generate malicious behavior even under constrainedconditions. Via ablative studies, we analyze the impact of different attackdesign choices followed by an evaluation of existing defence mechanisms againstthe proposed attack.

Adversarial Backdoor Attack by Naturalistic Data Poisoning on Trajectory Prediction in Autonomous Driving