EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Natural Occlusion-Based Backdoor Attacks: A Novel Approach to Compromising Pedestrian Detectors

Qiong Li iD, Yalun Wu iD, Qihuan Li iD, Xiaoshu Cui iD, Yuanwan Chen iD, Xiaolin Chang iD, Jiqiang Liu iD, Wenjia Niu iD

Sensors · 2025

DOI: 10.3390/s25134203 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, 2025. Verfasst von Qiong Li und 7 weiteren Autor:innen. Beteiligte Einrichtung: School of Cyberspace Science and Technology, Beijing Jiaotong University, Beijing 100044, China. Hauptthemen laut MEDLINE: Pedestrians. Die Arbeit nennt 64 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Fundamental Research Funds for the Central Universities, Central Funds Guiding the Local Science and Technology Development, National Natural Science Foundation of China.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Pedestrian detection systems are widely used in safety-critical domains such as autonomous driving, where deep neural networks accurately perceive individuals and distinguish them from other objects. However, their vulnerability to backdoor attacks remains understudied. Existing backdoor attacks, relying on unnatural digital perturbations or explicit patches, are difficult to deploy stealthily in the physical world. In this paper, we propose a novel backdoor attack method that leverages real-world occlusions (e.g., backpacks) as natural triggers for the first time. We design a dynamically optimized heuristic-based strategy to adaptively adjust the trigger’s position and size for diverse occlusion scenarios, and develop three model-independent trigger embedding mechanisms for attack implementation. We conduct extensive experiments on two different pedestrian detection models using publicly available datasets. The results demonstrate that while maintaining baseline performance, the backdoored models achieve average attack success rates of 75.1% on KITTI and 97.1% on CityPersons datasets, respectively. Physical tests verify that pedestrians wearing backpack triggers could successfully evade detection under varying shooting distances of iPhone cameras, though the attack failed when pedestrians rotated by 90°, confirming the practical feasibility of our method. Through ablation studies, we further investigate the impact of key parameters such as trigger patterns and poisoning rates on attack effectiveness. Finally, we evaluate the defense resistance capability of our proposed method. This study reveals that common occlusion phenomena can serve as backdoor carriers, providing critical insights for designing physically robust pedestrian detection systems.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

PedestriansHumansAlgorithmsNeural Networks, Computer

Stichwörter der Autor:innen

backdoor attackdeep neural networksocclusion triggerpedestrian detection

Autor:innen und Institutionen

  • School of Cyberspace Science and Technology, Beijing Jiaotong University, Beijing 100044, China.
  • Beijing Key Laboratory of Security and Privacy in Intelligent Transportation, Beijing Jiaotong University, Beijing 100044, China.
  • School of Computer Science and Technology, Beijing Jiaotong University, Beijing 100044, China.

ORCID-Profile: Qiong Li, Yalun Wu, Qihuan Li, Xiaoshu Cui, Yuanwan Chen, Xiaolin Chang, Jiqiang Liu, Wenjia Niu

Förderung

  • Fundamental Research Funds for the Central Universities
  • Central Funds Guiding the Local Science and Technology Development
  • National Natural Science Foundation of China
  • Open Competition Mechanism to Select the Best Candidates in Shijiazhuang, Hebei Province, China

Zitieren

APA

Li, Q., Wu, Y., Li, Q., Cui, X., Chen, Y., Chang, X., Liu, J., & Niu, W. (2025). Natural Occlusion-Based Backdoor Attacks: A Novel Approach to Compromising Pedestrian Detectors. Sensors. https://doi.org/10.3390/s25134203

Vancouver

Li Q, Wu Y, Li Q, Cui X, Chen Y, Chang X, et al. Natural Occlusion-Based Backdoor Attacks: A Novel Approach to Compromising Pedestrian Detectors. Sensors. 2025. doi:10.3390/s25134203

BibTeX
@article{Li2025,
  author = {Li, Qiong and Wu, Yalun and Li, Qihuan and Cui, Xiaoshu and Chen, Yuanwan and Chang, Xiaolin and Liu, Jiqiang and Niu, Wenjia},
  title = {Natural Occlusion-Based Backdoor Attacks: A Novel Approach to Compromising Pedestrian Detectors},
  journal = {Sensors},
  year = {2025},
  doi = {10.3390/s25134203}
}
Kostenlos starten →