High-Resolution Remote Sensing Image Integrity Authentication Method Considering Both Global And Local Features

Remote sensing imagery is an important achievement of human earth observation,which provides a new perspective of geological observation "seeing the earth from a spatial perspective".At present,remote sensing images have been applied to many fields such as environment,geology and ecology,which greatly support the development of modern geography and other related disciplines.With the emergence of information sharing services,remote sensing images are widely disseminated,distributed and used in the Internet environment.The wide application of remote sensing image processing software(such as ENVI,ERDAS,etc.)also exposes remote sensing images to the possibility of being modified intentionally or unintentionally.From the perspective of information security,how to verify the content integrity of remote sensing data in the process of transmission and application,verify the authenticity of the data content and ensure that it is not maliciously tampered with,is an important issue faced by data distribution and managers,and also the basic guarantee for the safe dissemination and use of remote sensing data in the Internet environment.Integrity authentication refers to the process of verifying whether the content information is changed during the process of data transmission and application.The existing remote sensing image integrity authentication methods are mainly based on digital signature technology and fragile watermarking technology.The remote sensing image integrity authentication algorithm based on digital signature technology and fragile watermarking technology is essentially bitlevel authentication,i.e.,the data is recognized as tampered data as long as a 1-bit change occurs.This is less usable in practical applications because some content-preserving operations(e.g.,copyright watermark embedding,format conversion,etc.)do not change the content information of the data,yet they are susceptible to be identified as tampering by the above methods.Perceptual hashing,a common content retrieval and matching technique for multimedia data,is characterized by its robustness to content-retentive operations and sensitivity to contenttampering operations.There have been many papers applying perceptual hashing techniques to remote sensing data integrity authentication.However,in summary,there are three problems in the existing studies:(1)there is no precise definition of "remote sensing data integrity",i.e.,the boundary between tampered and non-tampered images is blurred;(2)Existing algorithms mostly use one image feature(such as feature points,edge lines,texture information,etc.)to construct a hash sequence,which is not universal for image authentication under mixed themes;(3)In the Internet sharing and dissemination environment,the improvement of time efficiency brings about the improvement of information sharing efficiency,so there is still room for improvement in the time efficiency of existing algorithms.In response to these problems,this paper adopts the idea of combining global and local features to construct a perceptual hash series suitable for remote sensing images.In general,the accuracy and interpretability of authentication have been improved,and the basic problem of tampering with boundaries has been studied in detail.The specific research content is as follows:(1)The characteristics of remote sensing images and the problems of their existing integrity authentication algorithms are analyzed;the principles,characteristics,and applicable scenarios of perceptual hashing are analyzed;various global feature extraction methods and local feature extraction methods are analyzed,and the possibility of combining global and local features for hash sequence construction is analyzed.(2)Aiming at the problem that the aforementioned remote sensing image perception hashing algorithms mostly use one kind of features for hash sequence construction,a highresolution remote sensing image integrity authentication algorithm combining global and local features is proposed.Zernike moments are used as the global features of the image,and FAST feature points are used as the local features of the image,and the two are combined for the construction of the hash sequence.This method achieves tampering localization at the sub-grid level and improves the time efficiency compared with the existing algorithms.(3)To address the aforementioned problem that the boundary between tampered and nontampered images is blurred,three datasets,namely,perceptually similar,tampered images and different images,are constructed.The analysis of remote sensing image tampering boundary is realized,that is,a more detailed analysis of "what is remote sensing image tampering" is realized.This analysis method is not only applicable to this paper,but also can provide a more feasible idea for the determination of the algorithm threshold in the subsequent related research.(4)DEM is a secondary product of high-resolution remote sensing images.Compared with high-resolution images,the integrity certification of DEM data poses a higher challenge to the design of the algorithm.Based on this,this paper proposes a perceptual hash integrity authentication algorithm combining DCT transform and geostatistical metric for aerial remote sensing DEM data.The algorithm overcomes the problem that existing authentication algorithms are not robust to content-preserving operations and has the function of tampering localization.