科研速览 · Science Skim继续刷下去 · Keep skimming →
◇ arXiv2026-08-19· cs.CV

Detecting Backdoors in Object Detection via Pre-NMS Prediction Distribution Shift

Longtian Wang, Zhengyu Zhao, Chenhao Lin, Le Yang, Shiwei Wang, Yuhan Zhi, Xiaofei Xie, Chao Shen

原始摘要(英文原文)· Original abstract
Object detection models deployed in safety-critical applications remain vulnerable to backdoor attacks that cause targeted misbehaviors when a hidden trigger is present. Existing detection methods either rely on trigger inversion or exploit architecture-specific assumptions, and critically, representative existing methods fail to generalize reliably to scene-level attacks, where a single trigger induces anomalous behavior across all objects in the scene simultaneously. We present DistScan, a backdoor detection framework based on a simple but previously unexploited observation: backdoor injection systematically shifts a model's pre-NMS prediction class distribution away from its training class frequencies, even on clean inputs without any trigger present. DistScan aggregates intermediate class predictions over a clean validation set and flags a model as backdoored if the resulting distribution deviates significantly from the training class frequencies, requiring no model weight access, no trigger knowledge, and no additional training. Extensive experiments on MS-COCO and PASCAL VOC across two architectures and three scene-level attack scenarios demonstrate that DistScan substantially outperforms existing methods, improving average detection accuracy over the best-performing applicable baseline by 27.32 percentage points.
读原文 · Read the paper ↗

AI 追问PRO

登录后使用 AI 追问

讨论区

登录后参与讨论

相关论文 · Related

Detecting Backdoors in Object Detection via Pre-NMS Prediction Distribution Shift — 科研速览 Science Skim