Research map / Adversarial robustness and security / Black-box and transfer attacks
Adversarial Attack on Deep Cross-Modal Hamming Retrieval
Chao Li, Shangqian Gao, Cheng Deng, Wei Liu, Heng Huang
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 attack iterations from 100 to 500 (ablation) | AACH on DCMH · MIRFlickr-25K I->T 32-bit | MAP: 0.654 -> 0.631 | Improved |
| increase attack iterations from 100 to 500 (ablation) | AACH on SSAH · NUS-WIDE I->T 32-bit | MAP: 0.462 -> 0.395 | Improved |
| increase attack iterations from 100 to 500 (ablation) | AACH on CMHH · MS COCO I->T 32-bit | MAP: 0.537 -> 0.411 | Improved |
| increase attack iterations from 100 to 500 (ablation) | AACH on CMHH · MIRFlickr-25K T->I 32-bit | MAP: 0.599 -> 0.592 | No clear effect |
| increase adversarial queries from 200 to 500 (ablation) | AACH on SSAH · MIRFlickr-25K I->T 32-bit | MAP: 0.599 -> 0.564 | Improved |
| increase adversarial queries from 200 to 500 (ablation) | AACH on DCMH · MS COCO T->I 32-bit | MAP: 0.540 -> 0.475 | Improved |
| increase adversarial queries from 500 to 1000 (ablation) | AACH on DCMH · MS COCO I->T 32-bit | MAP: 0.461 -> 0.519 | Hurt |
| increase adversarial queries from 500 to 1000 (ablation) | AACH on SSAH · NUS-WIDE I->T 32-bit | MAP: 0.395 -> 0.376 | Improved |
| use image adversarial queries instead of text adversarial queries | AACH on DCMH · MIRFlickr-25K 32-bit | MAP: 0.641 -> 0.631 | Improved |
| attack with 32-bit surrogate-based examples against 16-bit target | AACH on DCMH · MIRFlickr-25K I->T | MAP: 0.752 -> 0.623 | Improved |
| attack with 32-bit surrogate-based examples against 64-bit target | AACH on SSAH · NUS-WIDE I->T | MAP: 0.659 -> 0.388 | Improved |
| use black-box AACH instead of white-box CMLA | CMLA · MIRFlickr-25K I->T 32-bit DCMH | MAP: 0.521 -> 0.631 | Hurt |
Closest papers by meaning
- Cross-Modal Learning with Adversarial Samples (NeurIPS 2019 · similarity 0.90)
- Self-Supervised Adversarial Hashing Networks for Cross-Modal Retrieval (CVPR 2018 · similarity 0.87)
- Mutual-Modality Adversarial Attack with Semantic Perturbation (AAAI 2024 · similarity 0.85)
- Once and for All: Universal Transferable Adversarial Perturbation against Deep Hashing-Based Facial Image Retrieval (AAAI 2024 · similarity 0.85)
- Robust Self-Paced Hashing for Cross-Modal Retrieval with Noisy Labels (AAAI 2025 · similarity 0.84)
- Dual Self-Paced Cross-Modal Hashing (AAAI 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.
