← Back to the feed
AI & softwareResearch findings

Evaluation of Autonomous Driving System Vulnerability via Dynamic Adversarial Attacks

Original published 2023.12.10 13:14 KSTAdded to Blackdeck 2026.11.15 08:28 KST

What has changed?

A novel attack method dynamically manipulates object detection by using non-co-located adversarial patches.

Key points

  1. Introduces a dynamic attack method that manipulates autonomous vehicle decision-making processes.
  2. Uses moving adversarial patches to mislead object detection, especially during intersection crossings and lane changes.
  3. Proposes SIT-Net to simulate environmental effects, reducing the gap between simulation and real-world conditions.

Connect it to your work

The attack uses moving, adaptive patches to achieve stealthier and more flexible deception.

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivDynamic Adversarial Attacks on Autonomous Driving Systems