Research map / Adversarial robustness and security / Attacks on text and graphs
Manipulating SGD with Data Ordering Attacks
I Shumailov, Zakhar Shumaylov, Dmitry Kazhdan, Yiren Zhao, Nicolas Papernot, Murat A Erdogdu, Ross Anderson
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 |
|---|---|---|---|
| Reshuffle individual data points within batches using oscillation outward | ResNet-18 · CIFAR-10 | test accuracy: 90.51 -> 26.13 | Hurt |
| Reorder individual data points within batches using oscillation outward | ResNet-50 · CIFAR-100 | test accuracy: 75.56 -> 18.00 | Hurt |
| Reshuffle individual data points within batches from high to low surrogate loss | Sparse mean EmbeddingBag model · AGNews | test accuracy: 90.87 -> 56.54 | Hurt |
| Reorder batches using oscillation inward | Sparse mean EmbeddingBag model · AGNews | test accuracy: 90.87 -> 91.10 | No clear effect |
| Use natural reordered batches to create a whitebox flag-like backdoor | VGG16 · CIFAR-10, batch size 32 | trigger accuracy: 10.91±1.53 -> 91.03 ± 12.96 | Improved |
| Use natural reordered batches with a surrogate model to create a blackbox flag-like backdoor | VGG16 · CIFAR-10, batch size 32 | trigger accuracy: 10.91±1.53 -> 56.31 ± 19.57 | Improved |
| Use natural reordered batches to create a whitebox nine-white-lines backdoor | VGG16 · CIFAR-10, batch size 32 | trigger accuracy: 10.91±1.53 -> 33.93 ± 7.37 | Improved |
| Attack training at epoch 10 with low-high batch reordering | ResNet-18 · CIFAR-10, epoch 100 | performance: -3% | Hurt |
Closest papers by meaning
- Rethinking Backdoor Attacks on Dataset Distillation: A Kernel Method Perspective (ICLR 2024 · similarity 0.85)
- Rethinking Adversarial Robustness in the Context of the Right to be Forgotten (ICML 2024 · similarity 0.84)
- Backdoor Attacks via Machine Unlearning (AAAI 2024 · similarity 0.83)
- Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond (ICML 2025 · similarity 0.83)
- Adversarial Perturbations Are Formed by Iteratively Learning Linear Combinations of the Right Singular Vectors of the Adversarial Jacobian (ICML 2025 · similarity 0.83)
- Online Learning in the Random-Order Model (ICML 2025 · similarity 0.83)
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.
