Research map / Adversarial robustness and security / Black-box and transfer attacks
You See What I Want You To See: Exploring Targeted Black-Box Transferability Attack for Hash-Based Image Retrieval Systems
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Generate perturbations that remain adversarial after Gaussian noise is added (NAG) | Targeted black-box attacks on HashNet image retrieval · ImageNet, vulnerable pairs | attack success rate: 16.85% vs. 11.33/11.61% (DI/DI-Mom), 11.05% (PGD), and 9.22% (AA) | Helped |
| Generate perturbations that remain adversarial after Gaussian noise is added (NAG) | Targeted black-box attacks on HashNet image retrieval · ImageNet, normal pairs | attack success rate: 1.82% vs. 0.516%, 0.546%, 0.87%, and 0.22% | Helped |
| Select vulnerable source-target pairs using white-box Hamming distance | Targeted black-box transfer attacks for deep hashing · ImageNet, vulnerable versus normal pairs | black-box attack success: mean around 16 vs.25 | Helped |
| Increase Gaussian noise added to adversarial inputs | White-box targeted attacks for HashNet · ImageNet, η∞=16 and η∞=32 | white-box attack success rate: PGD: 96.7 -> 4.5 (η∞=16); 100.0 -> 12.4 (η∞=32) | Hurt |
| Use random-noise tolerance to estimate adversarial-region size | Transferability prediction for deep hashing attacks · ImageNet, compared attack strategies | black-box transferability | Helped |
Closest papers by meaning
- Transfer Adversarial Hashing for Hamming Space Retrieval (AAAI 2018 · similarity 0.89)
- Once and for All: Universal Transferable Adversarial Perturbation against Deep Hashing-Based Facial Image Retrieval (AAAI 2024 · similarity 0.89)
- HUANG: A Robust Diffusion Model-based Targeted Adversarial Attack Against Deep Hashing Retrieval (AAAI 2025 · similarity 0.89)
- Black-Box Adversarial Attack with Transferable Model-based Embedding (ICLR 2020 · similarity 0.89)
- Evade Deep Image Retrieval by Stashing Private Images in the Hash Space (CVPR 2020 · similarity 0.88)
- Perturbing Across the Feature Hierarchy to Improve Standard and Strict Blackbox Attack Transferability (NeurIPS 2020 · similarity 0.88)
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.