Research map / Adversarial robustness and security / Attacks on text and graphs
SALSA VERDE: a machine learning attack on LWE with sparse small secrets
Cathy Yuanchen Li, Emily Wenger, Zeyuan Allen-Zhu, Francois Charton, Kristin E. Lauter
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 |
|---|---|---|---|
| reorder rows of the lattice matrix before reduction (ablation) | VERDE preprocessing · LWE n=256, log2 q=20 | reduction factor: 0.53 -> 0.47 | Improved |
| reduce BKZ floating point precision (ablation) | VERDE preprocessing · LWE n=256, log2 q=20 | preprocessing time (hrs/CPU): 136 -> 24 | Improved |
| lower omega from 15 to 10 (ablation) | VERDE preprocessing · LWE n=256, log2 q=20 | reduction factor: 0.47 -> 0.43 | Mixed |
| interleave BKZ 2.0 with efficient reduction, adaptive blocksize, early stopping (ablation) | VERDE preprocessing · LWE n=256, log2 q=20 | preprocessing time (hrs/CPU): 38 -> 7.5 | Improved |
| run distinguisher on held-out preprocessed vectors instead of random vectors (ablation) | VERDE secret recovery · n=256, log2 q=23, binary | fraction recovered: 1/5 -> 2/5 | Improved |
| use smaller base B=5363 instead of 7150 (ablation) | VERDE tokenization · n=256, log2 q=16, Gaussian h=6 | fraction recovered: 1/10 -> 3/10 | Improved |
| run 5 model initializations instead of 1 (ablation) | VERDE · n=256, log2 q=12, binary h=4 | fraction recovered: 1/10 -> 3/10 | Improved |
| increase preprocessing time from 7 to 35 days (ablation) | VERDE · n=512, log2 q=41, binary | highest h recovered: 20 -> 63 | Improved |
| lower modulus from log2 q=20 to 12 | VERDE · n=256, binary | highest h recovered: 33 -> 6 | Hurt |
| raise NoMod percentage of training data above threshold | VERDE · n=256, log2 q=20, binary | secret recovery: 67% | Improved |
| permute columns of A and re-preprocess | VERDE · n=256, log2 q=14, binary h=6-8 | secret recovery | Improved |
| replace seq2seq with encoder-only transformer with EMD loss | VERDE model · n=256 and 512 | epochs to recovery | Mixed |
Closest papers by meaning
- Making Hard Problems Easier with Custom Data Distributions and Loss Regularization: A Case Study in Modular Arithmetic (ICML 2025 · similarity 0.90)
- TAPAS: Datasets for Learning the Learning with Errors Problem (NeurIPS 2025 · similarity 0.88)
- Encryption-Friendly LLM Architecture (ICLR 2025 · similarity 0.82)
- BiShield-TEE: On the (In-)Security of Unilateral Weight Obfuscation in On-Device TEE-Shielded LLM Partition (NeurIPS 2026 · similarity 0.82)
- FicGCN: Unveiling the Homomorphic Encryption Efficiency from Irregular Graph Convolutional Networks (ICML 2025 · similarity 0.81)
- PrivDNFIS: Privacy-preserving and Efficient Deep Neuro-Fuzzy Inference System (AAAI 2025 · similarity 0.81)
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.
