Published in The 3rd USENIX Symposium on Vehicle Security and Privacy (VehicleSec '25), 2025
A zero-shot LLM-based reasoning framework for adversarial scenario analysis in autonomous vehicles, using formal traffic rules and chain-of-thought prompting.
Published in arXiv preprint arXiv:2506.11472, 2025
Demonstrates the inherent adversarial robustness of vision-language models for autonomous vehicle perception without adversarial training.
Published in CVPR 2025 – IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025
Proposes temporally coherent shadow-based adversarial attacks on traffic sign recognition systems, leveraging attention maps and evolutionary optimization.
Published in SAE Technical Paper, 2025
Evaluates vision-language models like LLaVA in post-crash analysis tasks, showing improved reasoning and generalization in AV-critical scenarios.
Published in 2025 29th International Computer Conference, Computer Society of Iran (CSICC), 2025
Presents Point-LN, a lightweight point cloud classification framework combining non-learnable geometry with a minimal trainable classifier.
Published in 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025
Introduces Point-GN, a non-parametric model using Gaussian positional encoding for efficient and scalable 3D point cloud classification.