Research map / Adversarial robustness and security / Black-box and transfer attacks
CamPI: Physical Adversarial Examples through Camera Power Signal Injection
Yanze Ren, Mingyuan Lv, Qinhong Jiang, Yan Jiang, Chen Yan, Xiaoyu Ji, Wenyuan Xu
Read the paperCheck 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 |
|---|---|---|---|
| increase CW trade-off coefficient c from 10^1 to 10^2 (ablation) | CamPI white-box CW-L2 · NIPS 2017 untargeted, 7-model average | attack success rate (%): 79.1 -> 95.5 (L2 4.0 -> 7.3) | Improved |
| increase c from 10^2 to 10^4 (ablation) | CamPI white-box CW-L2 · NIPS 2017 untargeted, 7-model average | attack success rate (%): 95.5 -> 96.8 (L2 7.3 -> 14.8) | No clear effect |
| relax L0 constraint from 1 to 256 perturbed blocks (ablation) | CamPI black-box DE-L0 · VGG-19 simulation | attack success rate (%): reaches 89.7% at 256 blocks | Improved |
| move from simulation to physical injection (white-box) | CamPI · 100 images, white-box | attack success rate (%): 98% -> 92% | Hurt |
| move from simulation to physical injection (black-box) | CamPI · 100 images, black-box L0=256 | attack success rate (%): 85% -> 82% | Hurt |
| add Expectation over Transformation to optimization | CamPI physical attack · physical robustness to brightness, rotation, crop, contrast, noise, translation | attack effectiveness (qualitative) | Improved |
| increase attack distance via longer power cable (1 m to 10 m) (ablation) | CamPI physical attack · physical cable length 1-10 m | induced amplitude and ASR | Mixed |
| target affinity class pairs instead of arbitrary targets | CamPI white-box targeted · ImageNet val | targeted attack success rate (%): 80% to 100% across listed pairs | Improved |
Closest papers by meaning
- Invisible Perturbations: Physical Adversarial Examples Exploiting the Rolling Shutter Effect (CVPR 2021 · similarity 0.90)
- Adversarial camera stickers: A physical camera-based attack on deep learning systems (ICML 2019 · similarity 0.88)
- Revisiting Adversarial Patches for Designing Camera-Agnostic Attacks against Person Detection (NeurIPS 2024 · similarity 0.84)
- Breaking Barriers in Physical-World Adversarial Examples: Improving Robustness and Transferability via Robust Feature (AAAI 2025 · similarity 0.84)
- MAGIC: Mastering Physical Adversarial Generation in Context Through Collaborative LLM Agents (AAAI 2026 · similarity 0.84)
- ProjAttacker: A Configurable Physical Adversarial Attack for Face Recognition via Projector (CVPR 2025 · similarity 0.84)
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.
