Research map / Adversarial robustness and security / Black-box and transfer attacks
A Large-Scale Multiple-Objective Method for Black-Box Attack against Object Detection
Siyuan Liang, Longkang Li, Yanbo Fan, Xiaojun Jia, Jingzhi Li, Baoyuan Wu, Xiaochun Cao
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 |
|---|---|---|---|
| replace the attack objective with TP+FP (minimize true positives and maximize false positives) (ablation) | PRFA · GFL | mAP after attack (lower is better): 0.31 -> 0.27 | Improved |
| replace the attack objective with TP+FP (ablation) | PRFA · YOLOX | mAP after attack (lower is better): 0.31 -> 0.28 | Improved |
| initialize with skip-based gradient-prior perturbation (ablation) | PRFA · GFL | mAP after attack (lower is better): 0.31 -> 0.21 | Improved |
| initialize with skip-based gradient-prior perturbation (ablation) | PRFA · YOLOX | mAP after attack (lower is better): 0.31 -> 0.31 | No clear effect |
| replace PRFA search with a genetic algorithm over the whole image (ablation) | PRFA with skip-based init · GFL | mAP after attack (lower is better): 0.21 -> 0.25 | Hurt |
| replace PRFA search with a genetic algorithm over the whole image (ablation) | PRFA with skip-based init · YOLOX | mAP after attack (lower is better): 0.31 -> 0.42 | Hurt |
| add random subset selection (ablation) | GA with skip-based init · GFL | mAP after attack (lower is better): 0.25 -> 0.20 | Improved |
| add random subset selection (ablation) | GA with skip-based init · YOLOX | mAP after attack (lower is better): 0.42 -> 0.29 | Improved |
| add divide-and-conquer (ablation) | GARS with skip-based init · GFL | mAP after attack (lower is better): 0.20 -> 0.18 | Improved |
| add divide-and-conquer (ablation) | GARS with skip-based init · YOLOX | mAP after attack (lower is better): 0.29 -> 0.31 | Hurt |
| add chain-based perturbations to form a mixed initial population (ablation) | GARSDC with skip-based init · GFL | average queries (lower is better): 3037 -> 1838 | Improved |
| add chain-based perturbations to form a mixed initial population (ablation) | GARSDC with skip-based init · YOLOX | mAP after attack (lower is better): 0.31 -> 0.23 | Improved |
Closest papers by meaning
- Blackbox Attacks via Surrogate Ensemble Search (NeurIPS 2022 · similarity 0.88)
- Parallel Rectangle Flip Attack: A Query-Based Black-Box Attack Against Object Detection (ICCV 2021 · similarity 0.87)
- Amnesia as a Catalyst for Enhancing Black Box Pixel Attacks in Image Classification and Object Detection (NeurIPS 2024 · similarity 0.85)
- Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior (ICML 2024 · similarity 0.85)
- Transferable Adversarial Attacks for Object Detection Using Object-Aware Significant Feature Distortion (AAAI 2024 · similarity 0.84)
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · 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.
