跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.23880 · v1

A Rank-One Bypass of Fixed Linear Erasure

首次提交 2026-11-18 13:32
版本更新 2026-11-18 13:32
雷达观测 2026-11-19 19:08
arXiv ↗PDF ↗

摘要

Linear concept erasure can make a target unreadable to a specified family of linear classifiers applied to a model's current internal features. In an open-weight release, however, an attacker can modify the layers that produce those features. We study a fixed affine erasure layer whose upstream model is allowed to change. If the layer transmits even one direction visible to an allowed downstream classifier, the attacker can write a target score into that direction and carry it through the layer. We show that this condition is also necessary, and that for a nonconstant linear target score the required change is a rank-one affine update. Across six CIFAR-10 class-pair tasks, the route restores nearly all pre-erasure target performance and can leave a deployed benign linear head's scores unchanged to numerical tolerance. We then price the attack: with eight target labels and 2,000 unlabelled calibration examples, a 10% mean relative feature edit yields 79.8% test accuracy, which is 91.2% of the fitted score's gain over chance; tighter edit limits sharply reduce success. Linear guardedness therefore remains a valid guarantee about the current representation, but it is not by itself a tamper-resistance guarantee for an editable open-weight model.

作者备注:Accepted at the NeurIPS 2026 Workshop on Child Safety in AI (poster). 10 pages, 3 figures

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-18 13:32 · 最新观测版本

A Rank-One Bypass of Fixed Linear Erasure

Linear concept erasure can make a target unreadable to a specified family of linear classifiers applied to a model's current internal features. In an open-weight release, however, an attacker can modify the layers that produce those features. We study a fixed affine erasure layer whose upstream model is allowed to change. If the layer transmits even one direction visible to an allowed downstream classifier, the attacker can write a target score into that direction and carry it through the layer. We show that this condition is also necessary, and that for a nonconstant linear target score the required change is a rank-one affine update. Across six CIFAR-10 class-pair tasks, the route restores nearly all pre-erasure target performance and can leave a deployed benign linear head's scores unchanged to numerical tolerance. We then price the attack: with eight target labels and 2,000 unlabelled calibration examples, a 10% mean relative feature edit yields 79.8% test accuracy, which is 91.2% of the fitted score's gain over chance; tighter edit limits sharply reduce success. Linear guardedness therefore remains a valid guarantee about the current representation, but it is not by itself a tamper-resistance guarantee for an editable open-weight model.

查看此版本解读 →查看此版本原文 ↗