AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Theory of Minimal Weight Perturbations in Deep Networks and its Applications for Low-Rank Activated Backdoor Attacks

Bethan Evans, Jared Tanner

ICML 2026 · Adversarial robustness and security › Backdoor and poisoning · approach: Backdoors in large multimodal models · 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
increase regularisation lambda to 200 when training the perturbation (ablation)Empirical single-layer weight perturbation · 5-layer network, 4-class syntheticclass flip success: flips at lambda=1 in all layers; none at lambda=200Hurt
perturb layers 1 to k jointly instead of a single layer (ablation)10-layer linear network · 4-class syntheticperturbation norm neededImproved
apply low-rank projection of the final layer to a backdoor trained with the precision-conditioned lossVGG16 on CIFAR-10 · CIFAR-10 rank 10 to rank 8ASR: 10.1 -> 96.2Improved
use precision-conditioned loss instead of control loss (ablation)VGG16 on CIFAR-10 · CIFAR-10 full precisionASR at full precision: 97.0 -> 10.1Improved
apply low-rank compression to Phi-2 fc2 layersPhi-2 backdoor · SQuADASR: 45.5 -> 86.0Improved
prune the backdoored ResNet18 to higher sparsity (ablation)ResNet18 on CIFAR-10 · single samplemargin bound satisfied / class flip: class flip first at 17% sparsityImproved

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