Exploring Adversarial Robustness of LiDAR Semantic Segmentation in Autonomous Driving
K. T. Yasas Mahima, Asanka Perera, Sreenatha Anavatti, Matt Garratt · 2023
Read Exploring Adversarial Robustness of LiDAR Semantic Segmentation in Autonomous Driving by K. T. Yasas Mahima, Asanka Perera, Sreenatha Anavatti, Matt Garratt (2023) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-10708872.txt.
Description
Mahima, Perera, Anavatti, and Garratt present a study from twenty twenty-three on trick attacks against laser sensing in self-driving cars. Laser scanners build a labeled map of the scene around a car. Small, hard-to-see changes to the input can fool the networks that label those maps. Such attacks are well studied for photos but little studied for laser maps. The team built three attack methods for point-based laser data. They tested them on networks trained on a known driving dataset.
One network proved most open to the attacks they tried. Ground-level points were the easiest to fool with small shifts. Attacks moved between networks unevenly, with point-based ones resisting better. The authors say the findings should guide stronger defenses. They call for more work on laser-specific attacks and guards.
Catalog record
| CREATOR | K. T. Yasas Mahima, Asanka Perera, Sreenatha Anavatti, Matt Garratt |
|---|---|
| YEAR | 2023 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |