AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Defending against Backdoor Attacks via Module Switching

Weijun Li, Ansh Arora, Xuanli He, Mark Dras, Qiongkai Xu

ICLR 2026 · Adversarial robustness and security › Backdoor and poisoning · approach: Trigger-based backdoor attacks · 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
switch modules between two backdoored models instead of averaging weightsRoBERTa-large · SST-2, BadNet + InsertSentASR avg: 31.9 -> 22.0Improved
use module switching instead of WAG on stealthier attack pairRoBERTa-large · SST-2, BadNet + LWSASR avg: 62.2 -> 40.4Improved
use module switching when merging benign with backdoored modelRoBERTa-large · SST-2, Benign + BadNetASR: 39.3 -> 12.2Improved
use module switching instead of WAG on vision pairsViT · CIFAR-10, BadNet + PhysicalBAASR avg: 39.6 -> 18.5Improved
use module switching instead of WAG/TIES on BadNet + WaNetViT · CIFAR-10, BadNet + WaNetASR avg: WAG 12.2, TIES 11.3, MSD 11.4No clear effect
use module switching when merging benign with PhysicalBA modelViT · CIFAR-10, Benign + PhysicalBAASR: 10.1 vs 10.1 (WAG)No clear effect
module switching vs WAG on clean accuracyRoBERTa-large · SST-2, BadNet + InsertSentCACC: 96.3 -> 96.2No clear effect
use search without early stopping (higher fitness score) (ablation)MSD strategy · SST-2ASR: -27.2% ASRImproved
add each of the first three heuristic rules (intra-layer, consecutive, residual) to the fitness (ablation)MSD · SST-2ASRImproved
use MSD with three backdoored models instead of WAGRoBERTa-large · three-model fusionASR avg: below 20%Improved
reduce clean validation data from 50 to 20 samples per class (ablation)MSD selection · SST-2ASR of selected candidateNo clear effect
apply MSD to CNNs (ResNet-18/50) instead of WAGResNet · CIFAR-10ASRMixed

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