Research map / Adversarial robustness and security / Black-box and transfer attacks
Backpropagating Linearly Improves Transferability of Adversarial Examples
Yiwen Guo, Qizhang Li, Hao Chen
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 linear backpropagation instead of the standard gradient | I-FGSM from VGG-19 · CIFAR-10, ε=0.03, WRN transfer | success rate: 64.32% -> 91.60% | Improved |
| Use linear backpropagation instead of the standard gradient | I-FGSM from ResNet-50 · ImageNet, ε=0.05, average victim transfer | success rate: 38.64% -> 75.53% | Improved |
| Add ILA to LinBP | I-FGSM from ResNet-50 · ImageNet, ε=0.05, average victim transfer | success rate: 75.53% -> 83.62% | Improved |
| Add SGM to LinBP+ILA | I-FGSM from ResNet-50 · ImageNet, ε=0.05, average victim transfer | success rate: 83.62% -> 84.72% | Improved |
| Add SGM to LinBP with λ=0.5 | I-FGSM from ResNet-50 · ImageNet, ε=0.05, average victim transfer | success rate: 75.53% -> 78.88% | Improved |
| Remove ReLUs in later blocks without fine-tuning (ablation) | I-FGSM from VGG-19 · CIFAR-10, ε=0.03, WRN transfer | success rate: 64.32% -> 76.70% | Improved |
| Remove ReLUs in later blocks without fine-tuning (ablation) | FGSM from VGG-19 · CIFAR-10, ε=0.03, WRN transfer | success rate: 39.52% -> 28.70% | Hurt |
| Fine-tune the ReLU-removed LinS model to its optimal epoch (ablation) | I-FGSM from VGG-19 · CIFAR-10, ε=0.03, WRN transfer | success rate: 76.70% -> 91.36% | Improved |
| Use linear backpropagation rather than the optimally fine-tuned LinS gradient (ablation) | I-FGSM from VGG-19 · CIFAR-10, ε=0.03, WRN transfer | success rate: 91.36% -> 91.60% | Improved |
| Continue fine-tuning the more-linear model beyond the most transferable point (ablation) | LinS source model · CIFAR-10 | transfer success rate | Hurt |
Closest papers by meaning
- A Theoretical View of Linear Backpropagation and its Convergence (TPAMI 2024 · similarity 0.92)
- Rethinking the Backward Propagation for Adversarial Transferability (NeurIPS 2023 · similarity 0.89)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.88)
- LRS: Enhancing Adversarial Transferability through Lipschitz Regularized Surrogate (AAAI 2024 · similarity 0.87)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.86)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 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.
