Research map / Adversarial robustness and security / Black-box and transfer attacks
Targeted Attack for Deep Hashing based Retrieval
Jiawang Bai, Bin Chen, Yiming Li, Dongxian Wu, Weiwei Guo, Shu-Tao Xia, En-Hui Yang
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 |
|---|---|---|---|
| use point-to-set anchor code with n_t=9 instead of point-to-point (n_t=1) target | DHTA · ImageNet (16 bits) | t-MAP (%): 44.35 -> 63.68 | Improved |
| use point-to-set anchor code instead of point-to-point target | DHTA · JHMDB (64 bits) | t-MAP (%): 44.38 -> 66.06 | Improved |
| use point-to-set anchor code instead of point-to-point target | DHTA · NUS-WIDE (HashNet target, 16 bits) | t-MAP (%): 78.10 -> 86.95 | Improved |
| increase object set size n_t beyond 1 (ablation) | DHTA · ImageNet and JHMDB | t-MAP (%) | Improved |
| increase n_t from 5 to 11 in open-set attack (ablation) | DHTA · ImageNet open-set (16 bits) | t-MAP (%): 33.67 -> 38.00 | Improved |
| increase attack iterations from 100 to 2000 (ablation) | DHTA · ImageNet (32 bits) | t-MAP (%): 66.29 -> 77.76 | Improved |
| increase attack iterations from 10 to 500 (ablation) | DHTA · JHMDB (16 bits) | t-MAP (%): 28.51 -> 56.47 | Improved |
| increase epsilon from 0.01 to 0.05 (ablation) | DHTA · ImageNet (32 bits) | t-MAP (%): 25.39 -> 79.17 | Improved |
| increase epsilon from 0.01 to 0.05 (ablation) | DHTA · JHMDB (32 bits) | t-MAP (%): 51.08 -> 62.76 | Improved |
| attack in Hamming space instead of feature space | Feature-based attack · ImageNet (16 bits) | t-MAP (%): 23.80 -> 63.68 | Improved |
| attack in Hamming space instead of feature space | Feature-based attack · JHMDB (64 bits) | t-MAP (%): 57.54 -> 66.06 | Mixed |
| use DHTA as a non-targeted attack versus HAG | HAG · ImageNet (16 bits) MAP | MAP (%): 1.88 -> 0.54 | Improved |
Closest papers by meaning
- Prototype-Supervised Adversarial Network for Targeted Attack of Deep Hashing (CVPR 2021 · similarity 0.90)
- You See What I Want You To See: Exploring Targeted Black-Box Transferability Attack for Hash-Based Image Retrieval Systems (CVPR 2021 · similarity 0.88)
- Deep Supervised Adversarial Robust Hashing for Retrieval (TPAMI 2026 · similarity 0.87)
- Once and for All: Universal Transferable Adversarial Perturbation against Deep Hashing-Based Facial Image Retrieval (AAAI 2024 · similarity 0.87)
- Deep Hash Distillation for Image Retrieval (ECCV 2022 · similarity 0.87)
- HUANG: A Robust Diffusion Model-based Targeted Adversarial Attack Against Deep Hashing Retrieval (AAAI 2025 · similarity 0.86)
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.
