Demo: ROI Attacks on Traffic Light Detection in High-Level Autonomous Driving

Kanglan Tang, Junjie Shen, Qi Alfred ChenSP Workshops 2021 · OpenReview

Abstract

To achieve high-level autonomous driving (AD), the AD system in an Autonomous Vehicle (AV) has two key modules: localization, which estimates the real-time position of the AV on a map, and perception, which incorporates sensors such as cameras, radars, and LiDARs [1] to perceive the surrounding environment of the AV. From high-level design, the localization algorithms and perception algorithms appear to be independent of each other. However, in this work, we discover an interesting design consideration that enables an attacker to blind or misguide the perception without tampering the perception sensor inputs themselves.
BibTeX
@inproceedings{tang2021demo,
  author = {Kanglan Tang and Junjie Shen and Qi Alfred Chen},
  title = {{Demo: ROI Attacks on Traffic Light Detection in High-Level Autonomous Driving}},
  booktitle = {SP Workshops},
  year = {2021},
  url = {https://openreview.net/forum?id=TQ9GA-nXDNdV},
}

Loading related content...

Figure