Research map / Adversarial robustness and security
Black-box and transfer attacks: research map
513 papers on Black-box and transfer attacks in Adversarial robustness and security, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 7 approaches. The busiest year so far is 2026.
What it is
Attacks on models the attacker cannot see inside: crafting adversarial examples by querying the model, or on a substitute model in the hope that they transfer to the real one.
Terms you will see on this page:
- Black-box attack
- an attack that only uses the model's outputs.
- Transferability
- adversarial examples made for one model also fooling another.
- Query efficiency
- how few model queries an attack needs.
Start reading with the most cited papers here:
- Boosting Adversarial Attacks With Momentum (CVPR 2018 · 3,047 citations)
- Ensemble Adversarial Training: Attacks and Defenses (ICLR 2018 · 2,314 citations)
Within Adversarial robustness and security, its share held steady from 11.4% in 2023–24 to 11.8% in 2025–26 (89 → 116 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Black-box and transfer attacks in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Data-free substitute training rising: 8% → 23% of the topic (7 → 27 papers)
- Transfer attacks on face recognition rising: 10% → 22% of the topic (9 → 25 papers)
- Query-efficient black-box attacks rising: 11% → 13% of the topic (10 → 15 papers)
- Transferable adversarial examples falling: 36% → 20% of the topic (32 → 23 papers)
- Robustness benchmarks and defenses falling: 22% → 15% of the topic (20 → 17 papers)
Papers from 2025–2026 that are already cited most:
- Chain of Attack: On the Robustness of Vision-Language Models Against Transfer-Based Adversarial Attacks (CVPR 2025 · 12 citations)
- Improving Adversarial Transferability on Vision Transformers via Forward Propagation Refinement (CVPR 2025 · 9 citations)
- Mind the Gap: Detecting Black-box Adversarial Attacks in the Making through Query Update Analysis (CVPR 2025 · 5 citations)
- From Head to Tail: Efficient Black-box Model Inversion Attack via Long-tailed Learning (CVPR 2025 · 4 citations)
Approaches and key papers
Query-based black-box attacks · 305 papers
Keywords: query, black box, box adversarial
Approaches in this cluster:
- Robustness benchmarks and defenses (114 papers)
Benchmark adversarial robustness and defend through architecture, denoising and loss design. - Query-efficient black-box attacks (106 papers)
Craft adversarial examples with limited queries via bandits, Bayesian optimization and priors. - Data-free substitute training (58 papers)
Attack black-box models by training substitutes without data or stealing models. - Decision-based boundary attacks (27 papers)
Attack models using only hard-label outputs by approximating decision boundaries.
Most cited and most cited since 2024:
- Ensemble Adversarial Training: Attacks and Defenses (ICLR 2018 · 2,314 citations)
- Defense Against Adversarial Attacks Using High-Level Representation Guided Denoiser (CVPR 2018 · 945 citations)
- DifAttack: Query-Efficient Black-Box Adversarial Attack via Disentangled Feature Space (AAAI 2024 · 22 citations)
- CosPGD: an efficient white-box adversarial attack for pixel-wise prediction tasks (ICML 2024 · 13 citations)
Transfer attacks · 208 papers
Keywords: transferable, adversarial transferability, transferability adversarial
Approaches in this cluster:
- Transferable adversarial examples (116 papers)
Boost transfer of adversarial examples across models with gradient and input-diversity techniques. - Transfer attacks on face recognition (79 papers)
Improve transfer attacks with momentum and checkpoints, including against face recognition. - Transfer attacks on vision transformers (13 papers)
Exploit ViT components such as tokens and attention gradients for transferable attacks.
Most cited and most cited since 2024:
- Boosting Adversarial Attacks With Momentum (CVPR 2018 · 3,047 citations)
- Improving Transferability of Adversarial Examples With Input Diversity (CVPR 2019 · 1,273 citations)
- Boosting Adversarial Transferability by Block Shuffle and Rotation (CVPR 2024 · 88 citations)
- DiffAM: Diffusion-based Adversarial Makeup Transfer for Facial Privacy Protection (CVPR 2024 · 47 citations)
Related topics in Adversarial robustness and security
- Certified and empirical robustness (1,356)
- Backdoor and poisoning (385)
- Attacks on text and graphs (1,300)
