Research map / Adversarial robustness and security / Attacks on text and graphs
PhysGAN: Generating Physical-World-Resilient Adversarial Examples for Autonomous Driving
Zelun Kong, Junfeng Guo, Ang Li, Cong Liu
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use PhysGAN instead of RP2 for the physical-world attack | Nvidia Dave-2 · Physical-world evaluation | steering angle MSE: 23.48 -> 73.94 | Improved |
| Use PhysGAN instead of RP2 for the physical-world attack | Udacity Cg23 · Physical-world evaluation | steering angle MSE: 25.15 -> 99.23 | Improved |
| Use PhysGAN instead of RP2 for the physical-world attack | Udacity Rambo · Physical-world evaluation | steering angle MSE: 32.54 -> 87.56 | Improved |
| Use PhysGAN rather than PhysFGSM, conditioning generation on extracted features from video slices | Nvidia Dave-2 · Dave Straight1 | steering angle MSE: 15.88 -> 106.69 | Improved |
| Use FGSM rather than PhysGAN, perturbing the entire image | Nvidia Dave-2 · Dave Straight1 | steering angle MSE: 106.69 -> 115.91 | Improved |
| Add the target-model feature encoder to the generator (ablation) | PhysGAN · Digital driving scenes | attack efficacy | Improved |
| Use a 3D video-slice tensor rather than a single middle frame to generate the adversarial sign (ablation) | PhysGAN · Continuous drive-by attack | steering angle error | Improved |
| Replace the original roadside sign with the PhysGAN adversarial sign | Nvidia Dave-2 · Online driving test | time-to-curb: 10s | Improved |
Closest papers by meaning
- PhysPatch: A Physically Realizable and Transferable Adversarial Patch Attack for Multimodal Large Language Models-based Autonomous Driving Systems (AAAI 2026 · similarity 0.88)
- Robust Physical-World Attacks on Deep Learning Visual Classification (CVPR 2018 · similarity 0.87)
- SlowTrack: Increasing the Latency of Camera-Based Perception in Autonomous Driving Using Adversarial Examples (AAAI 2024 · similarity 0.86)
- PhyCamo: A Robust Physical Camouflage via Contrastive Learning for Multi-View Physical Adversarial Attack (AAAI 2025 · similarity 0.85)
- Adversarial Backdoor Attack by Naturalistic Data Poisoning on Trajectory Prediction in Autonomous Driving (CVPR 2024 · similarity 0.85)
- MAGIC: Mastering Physical Adversarial Generation in Context Through Collaborative LLM Agents (AAAI 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
