AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Improving ML Attacks on LWE with Data Repetition and Stepwise Regression

Alberto Alfarano, Eshika Saxena, Emily Wenger, Francois Charton, Kristin E. Lauter

ICML 2026 · Adversarial robustness and security › Certified and empirical robustness · approach: Robust overfitting and regularization · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add data repetition and larger training setstransformer LWE attack · n=64 log2 q=20 unreduced, h=4-6secret recoveryImproved
add repetition (small data budget) (ablation)transformer LWE attack · n=256 log2 q=12 BKZ, 4M sampleshighest h/cruel bits: 10/3 -> 14/5Improved
add repetition (15x vs 1x) (ablation)transformer LWE attack · n=256 log2 q=20 BKZ, 4M sampleshighest h/cruel bits: 55/6 -> 60/8Improved
replace linear regression with stepwise regression for cool bits (ablation)cool bit recovery · n=256 log2 q=20, 38 cool bits, 20Msecrets recovered out of 20: 1 -> 11Improved
add dual stepwise regression on top of stepwise (ablation)cool bit recovery · n=256 log2 q=20, 45 cool bits, 20Msecrets recovered out of 20: 3 -> 18Improved
use dual stepwise regression instead of linear regression (ablation)cool bit recovery · n=512 log2 q=41, 80 cool bits, 1Msecrets recovered out of 20: 0 -> 19Improved
use LASSO or Ridge instead of linear regression (ablation)cool bit recovery · cool bit recoverysecrets recoveredNo clear effect
add cool/cruel region embedding (ablation)transformer input · n=256 log2 q=20, 400M samplessecrets with 4 cruel bits recovered: 2/5 -> 5/5Improved
train on synthetic instead of BKZ-reduced data (ablation)transformer LWE attack · all four settingshighest h/cruel bitsNo clear effect
increase model size (embedding 256-1024, 4-12 layers) (ablation)transformer LWE attack · n=256 log2 q=20, h=60/65/70, 100M datamodel-based attempts ANo clear effect
use penalty alpha=0.1 beta=0.1 instead of alpha=0 beta=1 (ablation)loss function · n=256 log2 q=20, 400M 2xhighest h/cruel bits: 65/8 -> 70/8Improved
add repetition (ablation)transformer LWE attack · n=256 log2 q=20scaling exponent alpha_R: 0.70 -> 1.31Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy