Deep Learning on Digital Image Splicing Detection Using CFA Artifacts

Int. J. Sociotechnology Knowl. Dev. Pub Date : 2020-04-01 DOI:10.4018/ijskd.2020040102

Nadheer Younus Hussien, R. Mahmoud, Hala H. Zayed

{"title":"Deep Learning on Digital Image Splicing Detection Using CFA Artifacts","authors":"Nadheer Younus Hussien, R. Mahmoud, Hala H. Zayed","doi":"10.4018/ijskd.2020040102","DOIUrl":null,"url":null,"abstract":"Digitalimageforgeryisaseriousproblemofanincreasingattentionfromtheresearchsociety.Image splicingisawell-knowntypeofdigitalimageforgeryinwhichtheforgedimageissynthesizedfrom twoormoreimages.Splicingforgerydetectionismorechallengingwhencomparedwithotherforgery typesbecausetheforgedimagedoesnotcontainanyduplicatedregions.Inaddition,unavailabilityof sourceimagesintroducesnoevidenceabouttheforgeryprocess.Inthisstudy,anautomatedimage splicingforgerydetectionschemeispresented.Itdependsonextractingthefeatureofimagesbased ontheanalysisofcolorfilterarray(CFA).Afeaturereductionprocessisperformedusingprincipal componentanalysis (PCA) to reduce thedimensionalityof the resulting featurevectors.Adeep beliefnetwork-basedclassifierisbuiltandtrainedtoclassifythetestedimagesasauthenticorspliced images.TheproposedschemeisevaluatedthroughasetofexperimentsonColumbiaImageSplicing DetectionEvaluationDataset(CISDED)underdifferentscenariosincludingaddingpostprocessing onthesplicedimagessuchJPEGcompressionandGaussianNoise.Theobtainedresultsrevealthat theproposedschemeexhibitsapromisingperformancewith95.05%precision,94.05%recall,94.05% truepositiverate,and98.197%accuracy.Moreover,theobtainedresultsshowthesuperiorityofthe proposedschemecomparedtootherrecentsplicingdetectionmethod. KeywoRDS Color Filter Array, Deep Belief Network, Deep Learning, Digital Image Forgery, Splicing Forgery","PeriodicalId":13656,"journal":{"name":"Int. J. Sociotechnology Knowl. Dev.","volume":"98 1","pages":"31-44"},"PeriodicalIF":0.0000,"publicationDate":"2020-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"12","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Int. J. Sociotechnology Knowl. Dev.","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.4018/ijskd.2020040102","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 12

Abstract

Digitalimageforgeryisaseriousproblemofanincreasingattentionfromtheresearchsociety.Image splicingisawell-knowntypeofdigitalimageforgeryinwhichtheforgedimageissynthesizedfrom twoormoreimages.Splicingforgerydetectionismorechallengingwhencomparedwithotherforgery typesbecausetheforgedimagedoesnotcontainanyduplicatedregions.Inaddition,unavailabilityof sourceimagesintroducesnoevidenceabouttheforgeryprocess.Inthisstudy,anautomatedimage splicingforgerydetectionschemeispresented.Itdependsonextractingthefeatureofimagesbased ontheanalysisofcolorfilterarray(CFA).Afeaturereductionprocessisperformedusingprincipal componentanalysis (PCA) to reduce thedimensionalityof the resulting featurevectors.Adeep beliefnetwork-basedclassifierisbuiltandtrainedtoclassifythetestedimagesasauthenticorspliced images.TheproposedschemeisevaluatedthroughasetofexperimentsonColumbiaImageSplicing DetectionEvaluationDataset(CISDED)underdifferentscenariosincludingaddingpostprocessing onthesplicedimagessuchJPEGcompressionandGaussianNoise.Theobtainedresultsrevealthat theproposedschemeexhibitsapromisingperformancewith95.05%precision,94.05%recall,94.05% truepositiverate,and98.197%accuracy.Moreover,theobtainedresultsshowthesuperiorityofthe proposedschemecomparedtootherrecentsplicingdetectionmethod. KeywoRDS Color Filter Array, Deep Belief Network, Deep Learning, Digital Image Forgery, Splicing Forgery

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

基于CFA伪影的深度学习数字图像拼接检测

Digitalimageforgeryisaseriousproblemofanincreasingattentionfromtheresearchsociety。Image splicingisawell-knowntypeofdigitalimageforgeryinwhichtheforgedimageissynthesizedfrom twoormoreimages。Splicingforgerydetectionismorechallengingwhencomparedwithotherforgery typesbecausetheforgedimagedoesnotcontainanyduplicatedregions。Inaddition，unavailabilityof sourceimagesintroducesnoevidenceabouttheforgeryprocess。Inthisstudy，anautomatedimage splicingforgerydetectionschemeispresented。Itdependsonextractingthefeatureofimagesbased ontheanalysisofcolorfilterarray(CFA)。Afeaturereductionprocessisperformedusingprincipal componentanalysis (PCA)→reduce→thedimensionalityof→结果→featurevectors。Adeep beliefnetwork-basedclassifierisbuiltandtrainedtoclassifythetestedimagesasauthenticorspliced images。TheproposedschemeisevaluatedthroughasetofexperimentsonColumbiaImageSplicing DetectionEvaluationDataset(CISDED)underdifferentscenariosincludingaddingpostprocessing onthesplicedimagessuchJPEGcompressionandGaussianNoise。Theobtainedresultsrevealthat theproposedschemeexhibitsapromisingperformancewith95.05%precision，94.05%recall，94.05% truepositiverate，and98.197%accuracy。Moreover，theobtainedresultsshowthesuperiorityofthe proposedschemecomparedtootherrecentsplicingdetectionmethod。关键词:彩色滤波器阵列，深度信念网络，深度学习，数字图像伪造，拼接伪造

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文去求助

来源期刊

Int. J. Sociotechnology Knowl. Dev.

自引率

0.00%

发文量