AAU-Net: Attention-Based Asymmetric U-Net for Subject-Sensitive Hashing of Remote Sensing Images

Kaimeng Ding,Jin Tian,Yueming Liu,Yu Wang,Shiping Chen,Yue Zeng

doi:10.3390/rs13245109

Abstract

The prerequisite for the use of remote sensing images is that their security must be guaranteed. As a special subset of perceptual hashing, subject-sensitive hashing overcomes the shortcomings of the existing perceptual hashing that cannot distinguish between “subject-related tampering” and “subject-unrelated tampering” of remote sensing images. However, the existing subject-sensitive hashing still has a large deficiency in robustness. In this paper, we propose a novel attention-based asymmetric U-Net (AAU-Net) for the subject-sensitive hashing of remote sensing (RS) images. Our AAU-Net demonstrates obvious asymmetric structure characteristics, which is important to improve the robustness of features by combining the attention mechanism and the characteristics of subject-sensitive hashing. On the basis of AAU-Net, a subject-sensitive hashing algorithm is developed to integrate the features of various bands of RS images. Our experimental results show that our AAU-Net-based subject-sensitive hashing algorithm is more robust than the existing deep learning models such as Attention U-Net and MUM-Net, and its tampering sensitivity remains at the same level as that of Attention U-Net and MUM-Net.

Highlights

With the development of Earth observation (EO) technology, remote sensing (RS) images have been widely used in many fields, such as environmental science [1,2], urban planning [3,4], disaster monitoring [5,6], agriculture [7,8], and surveying and mapping [9,10].the use of RS images has an implicit premise, i.e., the security of the RS image must be guaranteed
In the training process of attention-based asymmetric U-Net (AAU-Net), the batch size was set to 4, and the number of epochs was set to 100; ReLu and sigmoid were selected as the activation functions of the attention gate of AAU-Net, since we found through experimental comparison that if δ1 is ReLU and δ2 is sigmoid, the tampering sensitivity and robustness of the algorithm are better than other activation functions
We introduced the attention mechanism into subject-sensitive hashing and proposed an attention-based asymmetric U-Net (AAU-Net)

Summary

Introduction

The use of RS images has an implicit premise, i.e., the security of the RS image must be guaranteed. If a user uses a tampered RS image, his/her analysis results obtained from the tampered RS image would either be not accurate enough, or incorrect, both of which may lead to a wrong and dangerous decision in some applications; if the user is not sure whether the RS image has been tampered with or not, the value of that image can be greatly reduced, or it even becomes useless.

Methods

Findings

Discussion

Conclusion