Defense against Adversarial Patch Attacks for Aerial Image Semantic Segmentation by Robust Feature Extraction
-
Published:2023-03-21
Issue:6
Volume:15
Page:1690
-
ISSN:2072-4292
-
Container-title:Remote Sensing
-
language:en
-
Short-container-title:Remote Sensing
Author:
Wang Zhen1ORCID, Wang Buhong1, Zhang Chuanlei2, Liu Yaohui3ORCID
Affiliation:
1. School of Information and Navigation, Air Force Engineering University, FengHao East Road, Xi’an 710082, China 2. School of Artificial Intelligence, Tianjin University of Science and Technology, Dagu South Road, Tianjin 300457, China 3. School of Surveying and Geo-Informatics, Shandong Jianzhu University, FengMing Road, Jinan 250101, China
Abstract
Deep learning (DL) models have recently been widely used in UAV aerial image semantic segmentation tasks and have achieved excellent performance. However, DL models are vulnerable to adversarial examples, which bring significant security risks to safety-critical systems. Existing research mainly focuses on solving digital attacks for aerial image semantic segmentation, but adversarial patches with physical attack attributes are more threatening than digital attacks. In this article, we systematically evaluate the threat of adversarial patches on the aerial image semantic segmentation task for the first time. To defend against adversarial patch attacks and obtain accurate semantic segmentation results, we construct a novel robust feature extraction network (RFENet). Based on the characteristics of aerial images and adversarial patches, RFENet designs a limited receptive field mechanism (LRFM), a spatial semantic enhancement module (SSEM), a boundary feature perception module (BFPM) and a global correlation encoder module (GCEM), respectively, to solve adversarial patch attacks from the DL model architecture design level. We discover that semantic features, shape features and global features contained in aerial images can significantly enhance the robustness of the DL model against patch attacks. Extensive experiments on three aerial image benchmark datasets demonstrate that the proposed RFENet has strong resistance to adversarial patch attacks compared with the existing state-of-the-art methods.
Funder
Natural Science Foundation of China National Natural Science Foundation of China Natural Science Foundation of Shandong Province Shandong Top Talent Special Foundation
Subject
General Earth and Planetary Sciences
Reference70 articles.
1. Russell, B.J., Soffer, R.J., Ientilucci, E.J., Kuester, M.A., Conran, D.N., Arroyo-Mora, J.P., Ochoa, T., Durell, C., and Holt, J. (2023). The Ground to Space CALibration Experiment (G-SCALE): Simultaneous Validation of UAV, Airborne, and Satellite Imagers for Earth Observation Using Specular Targets. Remote Sens., 15. 2. Tu, W., Hu, Z., Li, L., Cao, J., Jiang, J., Li, Q., and Li, Q. (2018). Portraying Urban Functional Zones by Coupling Remote Sensing Imagery and Human Sensing Data. Remote Sens., 10. 3. Zhang, Y., Guo, L., Wang, Z., Yu, Y., Liu, X., and Xu, F. (2020). Intelligent Ship Detection in Remote Sensing Images Based on Multi-Layer Convolutional Feature Fusion. Remote Sens., 12. 4. A Review of Machine Learning in Processing Remote Sensing Data for Mineral Exploration;Shirmard;Remote Sens. Environ.,2022 5. Analysis of Remote Sensing Imagery for Disaster Assessment Using Deep Learning: A Case Study of Flooding Event;Yang;Soft Comput.,2019
Cited by
2 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
|
|