AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Towards Stealthy and Effective Backdoor Attacks on Lane Detection: A Naturalistic Data Poisoning Approach

Yifan Liao, Yuxin Cao, Yedi Zhang, Wentao He, Yan Xiao, Xianglong Du, Zhiyong Huang, Jin Song Dong

CVPR 2026 · Adversarial robustness and security › Backdoor and poisoning · approach: Backdoors in large multimodal models · 3 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 gradient-heatmap guided trigger placement (ablation)DBALD (LaneATT) · CULane LDAASR: 80.97 -> 81.65No clear effect
add gradient-heatmap guided trigger placement (ablation)DBALD (LaneATT) · CULane LRAASR: 70.16 -> 73.45Improved
add gradient-heatmap guided trigger placement (ablation)DBALD (LaneATT) · CULane LOAASR: 68.52 -> 74.19Improved
add gradient-heatmap guided trigger placement (ablation)DBALD (RESA) · CULane LRAASR: 58.26 -> 65.15Improved
add gradient-heatmap guided trigger placement (ablation)DBALD (RESA) · CULane LOAASR: 69.37 -> 76.86Improved
add lane consistency loss to diffusion trigger generation (ablation)DBALD · CULane forensic detectionUniDetection score (lower = stealthier): 12.13 -> 2.46Improved
add environment consistency loss to diffusion trigger generation (ablation)DBALD · CULane forensic detectionUniDetection score (lower = stealthier): 14.51 -> 2.46Improved
use diffusion naturalistic trigger instead of BadLane textureDBALD · CULane avg ASR, LaneATTASR: 54.56 -> 76.41Improved
use DBALD vs LD-Attack on ADNetDBALD · TuSimple avg ASR, ADNetASR: 59.78 -> 80.66Improved
use DBALD vs LD-Attack on SCNNDBALD · TuSimple avg ASR, SCNNASR: 92.76 -> 93.54No clear effect
apply rain, blur or occlusion to physical poisoned imagesDBALD (LaneATT, cones) · Physical LOAASR: 57% -> 45/48/49%Hurt
fine-tune poisoned model on clean data for 50 epochs (defense)DBALD (LaneATT, LOA) · CULane LOAASR: 74.19% -> 63.68%Hurt

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