Research map / Adversarial robustness and security / Black-box and transfer attacks
Consistency-Sensitivity Guided Ensemble Black-Box Adversarial Attacks in Low-Dimensional Spaces
Jianhe Yuan, Zhihai He
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 |
|---|---|---|---|
| add sensitivity-guided local refinement after direct transfer attack (ablation) | CSEA, ResNet victim · CIFAR-10 untargeted | success rate: 90.9% -> 100.0% | Improved |
| add sensitivity-guided local refinement after direct transfer attack (ablation) | CSEA, DenseNet victim · CIFAR-10 untargeted | success rate: 96.1% -> 100.0% | Improved |
| order block flips by sensitivity map instead of random order (ablation) | local block refinement · CIFAR-10 untargeted | mean queries: 140.7 -> 113.3 | Improved |
| let the attacker keep memory of queries from earlier images | CSEA, ResNet victim · CIFAR-10 untargeted | mean queries: 74.2 -> 43.2 | Improved |
| let the attacker keep memory of queries from earlier images | CSEA, DenseNet victim · CIFAR-10 untargeted | mean queries: 65.4 -> 11.4 | Improved |
| let the attacker keep memory of queries from earlier images | CSEA, ResNet victim · CIFAR-10 targeted | mean queries: 930.4 -> 860.4 | Improved |
| use CSEA instead of Subspace Attack | Subspace Attack · ImageNet SENet untargeted | mean queries: 456 -> 360 | Improved |
| use CSEA instead of Subspace Attack | Subspace Attack · ImageNet PNAS-Net untargeted | mean queries: 680 -> 398 | Improved |
| use CSEA instead of Subspace Attack | Subspace Attack · ImageNet Inception-v3 untargeted | failure rate: 1.1% -> 1.8% | Hurt |
| use CSEA instead of NP-Attack | NP-Attack · CIFAR-10 targeted, ResNet victim | mean queries: 936 -> 930.4 | No clear effect |
| use CSEA instead of SimBA | SimBA · CIFAR-10 untargeted, ResNet victim | mean queries: 432.1 -> 74.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Subspace Attack — Subspace Attack: Exploiting Promising Subspaces for Query-Efficient Black-box Attacks (NeurIPS 2019) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Blackbox Attacks via Surrogate Ensemble Search (NeurIPS 2022 · similarity 0.90)
- Scaling Laws for Black-box Adversarial Attacks (ECCV 2026 · similarity 0.89)
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · similarity 0.89)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.87)
- Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior (ICML 2024 · similarity 0.87)
- BRUSLEATTACK: A QUERY-EFFICIENT SCORE- BASED BLACK-BOX SPARSE ADVERSARIAL ATTACK (ICLR 2024 · similarity 0.87)
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.
