Research map / Adversarial robustness and security / Attacks on text and graphs
Time Is All It Takes: Spike-Retiming Attacks on Event-Driven Spiking Neural Networks
Yi Yu, Qixin Zhang, Shuhan Ye, Xun Lin, Qianshan Wei, Kun Wang, Wenhan Yang, Dacheng Tao, Xudong Jiang
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 projected-in-the-loop (strict projection forward, soft shifter backward) (ablation) | PIL timing attack · DVS-Gesture binary B_1(8k) | ASR: 84.3 -> 98.5 | Improved |
| add projected-in-the-loop on integer grid B_inf(3) (ablation) | PIL timing attack · DVS-Gesture integer B_inf(3) | ASR: 63.0 -> 85.0 | Improved |
| add budget penalty R_p (ablation) | PIL timing attack · DVS-Gesture B_1 | ASR: binary 76.6 -> 98.5; integer 42.4 -> 79.1 | Improved |
| add capacity regularizer (ablation) | PIL timing attack · DVS-Gesture binary B_1(8k) | ASR: 98.5 -> 98.5 | No clear effect |
| add capacity regularizer on integer grid (ablation) | PIL timing attack · DVS-Gesture integer B_inf(3) | ASR: 77.6 -> 85.0 | Improved |
| use integer event grid instead of binary grid | SNN victims · DVS-Gesture ResNet18 B_1(8k) | ASR: 99.6 -> 65.4 (integer is more robust) | Improved |
| increase time bins T from 20 to 40 on binary grid (ablation) | PIL timing attack · DVS-Gesture binary B_1 | ASR: 90.6 -> 29.1 | Hurt |
| use targeted attack instead of untargeted (ablation) | PIL timing attack · DVS-Gesture binary B_inf(1) | ASR: 96.4 -> 26.8 | Hurt |
| use PLIF neurons instead of LIF (ablation) | victim VGGSNN · DVS-Gesture integer B_1(16k) | ASR: 79.1 -> 92.9 | Improved |
| adversarially train on integer grid with B_1(8k) | VGGSNN · DVS-Gesture | robust accuracy: clean about 69%; robust mid-50% to above 60% | Mixed |
| adversarially train on binary grid | VGGSNN · DVS-Gesture | clean accuracy: clean accuracy 22.92-48.26 vs 95.14 undefended | Hurt |
Closest papers by meaning
- Training Spiking Neural Networks with Event-driven Backpropagation (NeurIPS 2022 · similarity 0.89)
- Bullet Trains: Parallelizing Training of Temporally Precise Spiking Neural Networks (ICML 2026 · similarity 0.89)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.88)
- Advancing Training Efficiency of Deep Spiking Neural Networks through Rate-based Backpropagation (NeurIPS 2024 · similarity 0.88)
- Towards Effective and Sparse Adversarial Attack on Spiking Neural Networks via Breaking Invisible Surrogate Gradients (CVPR 2025 · similarity 0.87)
- Learning Delays in Spiking Neural Networks using Dilated Convolutions with Learnable Spacings (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.
