Comparative Analysis of Patch Attack on VLM-Based Autonomous Driving Architectures
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912957556850688 |
|---|---|
| author | Fernandez, David MohajerAnsari, Pedram Salarpour, Amir Cheng, Long Razi, Abolfazl Pesé, Mert D. |
| author_facet | Fernandez, David MohajerAnsari, Pedram Salarpour, Amir Cheng, Long Razi, Abolfazl Pesé, Mert D. |
| contents | Vision-language models are emerging for autonomous driving, yet their robustness to physical adversarial attacks remains unexplored. This paper presents a systematic framework for comparative adversarial evaluation across three VLM architectures: Dolphins, OmniDrive (Omni-L), and LeapVAD. Using black-box optimization with semantic homogenization for fair comparison, we evaluate physically realizable patch attacks in CARLA simulation. Results reveal severe vulnerabilities across all architectures, sustained multi-frame failures, and critical object detection degradation. Our analysis exposes distinct architectural vulnerability patterns, demonstrating that current VLM designs inadequately address adversarial threats in safety-critical autonomous driving applications. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_08897 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Comparative Analysis of Patch Attack on VLM-Based Autonomous Driving Architectures Fernandez, David MohajerAnsari, Pedram Salarpour, Amir Cheng, Long Razi, Abolfazl Pesé, Mert D. Computer Vision and Pattern Recognition Vision-language models are emerging for autonomous driving, yet their robustness to physical adversarial attacks remains unexplored. This paper presents a systematic framework for comparative adversarial evaluation across three VLM architectures: Dolphins, OmniDrive (Omni-L), and LeapVAD. Using black-box optimization with semantic homogenization for fair comparison, we evaluate physically realizable patch attacks in CARLA simulation. Results reveal severe vulnerabilities across all architectures, sustained multi-frame failures, and critical object detection degradation. Our analysis exposes distinct architectural vulnerability patterns, demonstrating that current VLM designs inadequately address adversarial threats in safety-critical autonomous driving applications. |
| title | Comparative Analysis of Patch Attack on VLM-Based Autonomous Driving Architectures |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.08897 |