MEGA Hub

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

Authors

Do you know Longtian Wang?You can claim authorship or link another user.Do you know Zhengyu Zhao?You can claim authorship or link another user.Do you know Chenhao Lin?You can claim authorship or link another user.Do you know Le Yang?You can claim authorship or link another user.Do you know Shiwei Wang?You can claim authorship or link another user.Do you know Yuhan Zhi?You can claim authorship or link another user.Do you know Xiaofei Xie?You can claim authorship or link another user.Do you know Chao Shen?You can claim authorship or link another user.

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.

Community

00