Summary of Sa-attack: Speed-adaptive Stealthy Adversarial Attack on Trajectory Prediction, by Huilin Yin et al.
SA-Attack: Speed-adaptive stealthy adversarial attack on trajectory prediction
by Huilin Yin, Jiaxiang Li, Pengju Zhen, Jun Yan
First submitted to arxiv on: 19 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
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 This paper proposes a novel adversarial attack method for trajectory prediction models used in autonomous vehicles. The approach, called SA-Attack, is designed to be stealthy and adaptable to realistic scenarios, unlike previous methods that focused on achieving high attack success rates without considering the practical implications. SA-Attack generates adversarial trajectories by incorporating information about forthcoming paths and using a vehicle-following method, while also reconstructing the trajectory from scratch to adapt to different speed scenarios. The method’s ability to fuse future trajectory trends and curvature constraints ensures smooth and stealthy attacks. The paper demonstrates the effectiveness of SA-Attack on nuScenes and Apolloscape datasets, showcasing its adaptability and stealthiness for various speed scenarios. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new way to trick machine learning models that predict car paths. This is important because these models are used in self-driving cars to plan safe routes. The current methods of tricking these models are not very good at hiding their tricks, which makes them easy to detect. The new method, called SA-Attack, can adapt to different speeds and hide its tricks well. It does this by using a method that follows the path of other cars and incorporating information about where those cars will go next. This helps create smooth and realistic fake paths that are hard to detect. The paper tests this method on two big datasets and shows it works well. |
Keywords
* Artificial intelligence * Machine learning
