APRICOT: A Dataset of Physical Adversarial Attacks on Object Detection
| Autor: | Elizabeth M. Merkhofer, Laura Strickhart, Nicole Lape, Keith Manville, Matthew Walmer, A. Braunegg, Amartya Chakraborty, Sara Leary, Michael Krumdick |
|---|---|
| Rok vydání: | 2020 |
| Předmět: | |
| Zdroj: | Computer Vision – ECCV 2020 ISBN: 9783030585884 ECCV (21) |
| DOI: | 10.1007/978-3-030-58589-1_3 |
| Popis: | Physical adversarial attacks threaten to fool object detection systems, but reproducible research on the real-world effectiveness of physical patches and how to defend against them requires a publicly available benchmark dataset. We present APRICOT, a collection of over 1,000 annotated photographs of printed adversarial patches in public locations. The patches target several object categories for three COCO-trained detection models, and the photos represent natural variation in position, distance, lighting conditions, and viewing angle. Our analysis suggests that maintaining adversarial robustness in uncontrolled settings is highly challenging but that it is still possible to produce targeted detections under white-box and sometimes black-box settings. We establish baselines for defending against adversarial patches via several methods, including using a detector supervised with synthetic data and using unsupervised methods such as kernel density estimation, Bayesian uncertainty, and reconstruction error. Our results suggest that adversarial patches can be effectively flagged, both in a high-knowledge, attack-specific scenario and in an unsupervised setting where patches are detected as anomalies in natural images. This dataset and the described experiments provide a benchmark for future research on the effectiveness of and defenses against physical adversarial objects in the wild. The APRICOT project page and dataset are available at apricot.mitre.org. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|