Summary of Towards Robust Physical-world Backdoor Attacks on Lane Detection, by Xinwei Zhang et al.
Towards Robust Physical-world Backdoor Attacks on Lane Detection
by Xinwei Zhang, Aishan Liu, Tianyuan Zhang, Siyuan Liang, Xianglong Liu
First submitted to arxiv on: 9 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Deep learning-based lane detection (LD) is a crucial component of autonomous driving systems, such as adaptive cruise control. However, LD is vulnerable to backdoor attacks. This paper proposes BadLANE, a dynamic scene adaptation backdoor attack for LD that can withstand changes in real-world dynamic scene factors. To address the challenges posed by changing driving perspectives, an amorphous trigger pattern composed of shapeless pixels is designed. This allows the backdoor to be activated by various forms or shapes of mud spots or pollution on the road or lens, enabling adaptation to changes in vehicle observation viewpoints during driving. A meta-learning framework is also proposed to train meta-generators tailored to different environmental conditions. These generators produce meta-triggers that incorporate diverse environmental information as the initialization of the trigger patterns for backdoor implantation, thus enabling adaptation to dynamic environments. The paper’s extensive experiments on various commonly used LD models in both digital and physical domains validate the effectiveness of BadLANE attacks, outperforming other baselines significantly (+25.15% on average in Attack Success Rate). |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper talks about how to make a special kind of attack work better for autonomous driving systems like adaptive cruise control. Right now, these systems are vulnerable to this type of attack called backdoor attacks. The problem is that the current methods don’t work well in real-world scenarios because they don’t consider things like changes in view or weather. To fix this, the authors propose a new way to make the attack work better using something called an amorphous trigger pattern. This lets the attack be activated by different shapes and forms of pollution on the road or camera lens. The authors also created a special framework to train machines that can learn from different environmental conditions like weather or lighting. They tested their approach with various models and found it worked significantly better than other methods (+25.15% on average). |
Keywords
» Artificial intelligence » Deep learning » Meta learning
