Research map / Adversarial robustness and security / Attacks on text and graphs
CP-FREEZER: Latency Attacks Against Vehicular Cooperative Perception
Chenyi Wang, Ruoyu Song, Raymond Muller, Jean-Philippe Monteuuis, Z. Berkay Celik, Jonathan Petit, Ryan Gerdes, Ming Li
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 |
|---|---|---|---|
| add spatial-temporal feature warping to the attack (ablation) | CP-FREEZER · OPV2V, RTX 4090, AttFusion | RoI-L: 63.91× -> 81.10× | Improved |
| add spatial-temporal feature warping to the attack (ablation) | CP-FREEZER · OPV2V, RTX 4090, Where2comm | RoI-L: 0.70× -> 48.30× | Improved |
| add spatial-temporal feature warping to the attack (ablation) | CP-FREEZER · OPV2V, RTX 4090, CoAlign | RoI-L: 98.74× -> 86.90× | Hurt |
| add spatial-temporal feature warping to the attack (ablation) | CP-FREEZER · OPV2V, RTX 4090, V2VAM | RoI-P: 9.97× -> 21.09× | Improved |
| use CP-FREEZER latency loss instead of PGD or prior-art single-vehicle latency losses | feature-level attack · OPV2V, AttFusion | RoI-P: 12.28× / 13.26× -> 46.19× | Improved |
| use PGD (accuracy-degrading) as a latency attack | Where2comm · OPV2V, RTX 2080 Super | RoI-L: -0.2× | Hurt |
| raise the pre-NMS confidence threshold up to 0.35 (ablation) | victim post-processing defense · OPV2V | RoI-P, RoI-L | No clear effect |
| vary the NMS IoU threshold (ablation) | victim post-processing defense · OPV2V, RTX 4090 | RoI-L | No clear effect |
| cap the maximum number of proposals below 1000 (ablation) | victim post-processing defense · OPV2V, RTX 4090 | RoI-L: up to 20× remaining | Mixed |
| adversarially train the CP models against CP-FREEZER for five epochs | CP models · OPV2V | mAP and RoI-L: V2VAM mAP 0.89, RoI-L 18.7× | Hurt |
| add ROBOSAC integrity defense | CP under CP-FREEZER · OPV2V, AttFusion | RoI-L: 81.10× -> 777.85× | Hurt |
| run attack on higher-end GPU (RTX 4090 vs 2080 Super) | CP-FREEZER · OPV2V, AttFusion | RoI-L: 89.4× -> 81.1× | No clear effect |
Closest papers by meaning
- CP-Guard: Malicious Agent Detection and Defense in Collaborative Bird’s Eye View Perception (AAAI 2025 · similarity 0.88)
- SlowTrack: Increasing the Latency of Camera-Based Perception in Autonomous Driving Using Adversarial Examples (AAAI 2024 · similarity 0.87)
- CoSDH: Communication-Efficient Collaborative Perception via Supply-Demand Awareness and Intermediate-Late Hybridization (CVPR 2025 · similarity 0.85)
- RCP-Bench: Benchmarking Robustness for Collaborative Perception Under Diverse Corruptions (CVPR 2025 · similarity 0.84)
- Privacy-Preserving V2X Collaborative Perception Integrating Unknown Collaborators (AAAI 2025 · similarity 0.84)
- STAMP: Scalable Task- And Model-agnostic Collaborative Perception (ICLR 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.
